pageserver - use batch_fsync for historics

2026-01-22 12:52:55 +00:00 · 2021-10-09 11:25:42 -07:00
86 changed files with 1346 additions and 1848 deletions
--- a/.circleci/config.yml
+++ b/.circleci/config.yml
@@ -241,21 +241,17 @@ jobs:
            if << parameters.run_in_parallel >>; then
              EXTRA_PARAMS="-n4 $EXTRA_PARAMS"
            fi;
-            ./netstat-script.sh &
-            NS_PID=$!
            # Run the tests.
            #
            # The junit.xml file allows CircleCI to display more fine-grained test information
            # in its "Tests" tab in the results page.
+            # -s prevents pytest from capturing output, which helps to see
+            # what's going on if the test hangs
            # --verbose prints name of each test (helpful when there are
            # multiple tests in one file)
            # -rA prints summary in the end
            # -n4 uses four processes to run tests via pytest-xdist
-            # -s is not used to prevent pytest from capturing output, because tests are running
-            # in parallel and logs are mixed between different tests
-            pipenv run pytest --junitxml=$TEST_OUTPUT/junit.xml --tb=short --verbose -rA $TEST_SELECTION $EXTRA_PARAMS
-            kill $NS_PID
-            awk '/===/ {if (count) print count; print; count=0; next} {count++} END {print count}' $TEST_OUTPUT/netstat.stdout > $TEST_OUTPUT/netstat_stats.stdout
+            pipenv run pytest --junitxml=$TEST_OUTPUT/junit.xml --tb=short -s --verbose -rA $TEST_SELECTION $EXTRA_PARAMS
      - run:
          # CircleCI artifacts are preserved one file at a time, so skipping
          # this step isn't a good idea. If you want to extract the
@@ -264,7 +260,7 @@ jobs:
          when: always
          command: |
            du -sh /tmp/test_output/*
-            find /tmp/test_output -type f ! -name "pg.log" ! -name "pageserver.log" ! -name "safekeeper.log" ! -name "regression.diffs" ! -name "junit.xml" ! -name "*.filediff" ! -name "*.stdout" ! -name "*.stderr" -delete
+            find /tmp/test_output -type f ! -name "pg.log" ! -name "pageserver.log" ! -name "wal_acceptor.log" ! -name "regression.diffs" ! -name "junit.xml" ! -name "*.filediff" ! -name "*.stdout" ! -name "*.stderr" -delete
            du -sh /tmp/test_output/*
      - store_artifacts:
          path: /tmp/test_output
@@ -329,7 +325,8 @@ jobs:
                \"inputs\": {
                  \"ci_job_name\": \"zenith-remote-ci\",
                  \"commit_hash\": \"$CIRCLE_SHA1\",
-                  \"remote_repo\": \"$LOCAL_REPO\"
+                  \"remote_repo\": \"$LOCAL_REPO\",
+                  \"zenith_image_branch\": \"$CIRCLE_BRANCH\"
                }
              }"

--- a/Cargo.lock
+++ b/Cargo.lock
@@ -26,21 +26,18 @@ dependencies = [
 "winapi",
 ]

-[[package]]
-name = "ansi_term"
-version = "0.12.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "d52a9bb7ec0cf484c551830a7ce27bd20d67eac647e1befb56b0be4ee39a55d2"
-dependencies = [
- "winapi",
-]
-
 [[package]]
 name = "anyhow"
 version = "1.0.42"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "595d3cfa7a60d4555cb5067b99f07142a08ea778de5cf993f7b75c7d8fabc486"

+[[package]]
+name = "arc-swap"
+version = "1.3.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e906254e445520903e7fc9da4f709886c84ae4bc4ddaf0e093188d66df4dc820"
+
 [[package]]
 name = "async-trait"
 version = "0.1.50"
@@ -301,7 +298,7 @@ version = "2.33.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "37e58ac78573c40708d45522f0d80fa2f01cc4f9b4e2bf749807255454312002"
 dependencies = [
- "ansi_term 0.11.0",
+ "ansi_term",
 "atty",
 "bitflags",
 "strsim",
@@ -390,6 +387,26 @@ dependencies = [
 "rustc_version",
 ]

+[[package]]
+name = "crossbeam-channel"
+version = "0.5.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "06ed27e177f16d65f0f0c22a213e17c696ace5dd64b14258b52f9417ccb52db4"
+dependencies = [
+ "cfg-if 1.0.0",
+ "crossbeam-utils",
+]
+
+[[package]]
+name = "crossbeam-utils"
+version = "0.8.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d82cfc11ce7f2c3faef78d8a684447b40d503d9681acebed6cb728d45940c4db"
+dependencies = [
+ "cfg-if 1.0.0",
+ "lazy_static",
+]
+
 [[package]]
 name = "crypto-mac"
 version = "0.10.0"
@@ -428,6 +445,16 @@ dependencies = [
 "dirs-sys",
 ]

+[[package]]
+name = "dirs-next"
+version = "2.0.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b98cf8ebf19c3d1b223e151f99a4f9f0690dca41414773390fc824184ac833e1"
+dependencies = [
+ "cfg-if 1.0.0",
+ "dirs-sys-next",
+]
+
 [[package]]
 name = "dirs-sys"
 version = "0.3.6"
@@ -439,6 +466,17 @@ dependencies = [
 "winapi",
 ]

+[[package]]
+name = "dirs-sys-next"
+version = "0.1.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "4ebda144c4fe02d1f7ea1a7d9641b6fc6b580adcfa024ae48797ecdeb6825b4d"
+dependencies = [
+ "libc",
+ "redox_users",
+ "winapi",
+]
+
 [[package]]
 name = "dlv-list"
 version = "0.2.3"
@@ -918,15 +956,6 @@ dependencies = [
 "cfg-if 1.0.0",
 ]

-[[package]]
-name = "matchers"
-version = "0.0.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f099785f7595cc4b4553a174ce30dd7589ef93391ff414dbb67f62392b9e0ce1"
-dependencies = [
- "regex-automata",
-]
-
 [[package]]
 name = "matches"
 version = "0.1.8"
@@ -1191,12 +1220,10 @@ dependencies = [
 "scopeguard",
 "serde",
 "serde_json",
- "signal-hook",
 "tar",
 "thiserror",
 "tokio",
 "toml",
- "tracing",
 "workspace_hack",
 "zenith_metrics",
 "zenith_utils",
@@ -1504,15 +1531,6 @@ dependencies = [
 "regex-syntax",
 ]

-[[package]]
-name = "regex-automata"
-version = "0.1.10"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "6c230d73fb8d8c1b9c0b3135c5142a8acee3a0558fb8db5cf1cb65f8d7862132"
-dependencies = [
- "regex-syntax",
-]
-
 [[package]]
 name = "regex-syntax"
 version = "0.6.25"
@@ -1671,6 +1689,12 @@ dependencies = [
 "webpki",
 ]

+[[package]]
+name = "rustversion"
+version = "1.0.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "61b3909d758bb75c79f23d4736fac9433868679d3ad2ea7a61e3c25cfda9a088"
+
 [[package]]
 name = "ryu"
 version = "1.0.5"
@@ -1828,32 +1852,12 @@ dependencies = [
 "opaque-debug",
 ]

-[[package]]
-name = "sharded-slab"
-version = "0.1.3"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "740223c51853f3145fe7c90360d2d4232f2b62e3449489c207eccde818979982"
-dependencies = [
- "lazy_static",
-]
-
 [[package]]
 name = "shlex"
 version = "1.0.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "42a568c8f2cd051a4d283bd6eb0343ac214c1b0f1ac19f93e1175b2dee38c73d"

-[[package]]
-name = "signal-hook"
-version = "0.3.10"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "9c98891d737e271a2954825ef19e46bd16bdb98e2746f2eec4f7a4ef7946efd1"
-dependencies = [
- "cc",
- "libc",
- "signal-hook-registry",
-]
-
 [[package]]
 name = "signal-hook-registry"
 version = "1.4.0"
@@ -1886,6 +1890,59 @@ version = "0.4.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "f173ac3d1a7e3b28003f40de0b5ce7fe2710f9b9dc3fc38664cebee46b3b6527"

+[[package]]
+name = "slog"
+version = "2.7.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8347046d4ebd943127157b94d63abb990fcf729dc4e9978927fdf4ac3c998d06"
+
+[[package]]
+name = "slog-async"
+version = "2.6.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c60813879f820c85dbc4eabf3269befe374591289019775898d56a81a804fbdc"
+dependencies = [
+ "crossbeam-channel",
+ "slog",
+ "take_mut",
+ "thread_local",
+]
+
+[[package]]
+name = "slog-scope"
+version = "4.4.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "2f95a4b4c3274cd2869549da82b57ccc930859bdbf5bcea0424bc5f140b3c786"
+dependencies = [
+ "arc-swap",
+ "lazy_static",
+ "slog",
+]
+
+[[package]]
+name = "slog-stdlog"
+version = "4.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8228ab7302adbf4fcb37e66f3cda78003feb521e7fd9e3847ec117a7784d0f5a"
+dependencies = [
+ "log",
+ "slog",
+ "slog-scope",
+]
+
+[[package]]
+name = "slog-term"
+version = "2.8.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "95c1e7e5aab61ced6006149ea772770b84a0d16ce0f7885def313e4829946d76"
+dependencies = [
+ "atty",
+ "chrono",
+ "slog",
+ "term",
+ "thread_local",
+]
+
 [[package]]
 name = "smallvec"
 version = "1.6.1"
@@ -1941,6 +1998,12 @@ dependencies = [
 "unicode-xid",
 ]

+[[package]]
+name = "take_mut"
+version = "0.2.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f764005d11ee5f36500a149ace24e00e3da98b0158b3e2d53a7495660d3f4d60"
+
 [[package]]
 name = "tap"
 version = "1.0.1"
@@ -1972,6 +2035,17 @@ dependencies = [
 "winapi",
 ]

+[[package]]
+name = "term"
+version = "0.7.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c59df8ac95d96ff9bede18eb7300b0fda5e5d8d90960e76f8e14ae765eedbf1f"
+dependencies = [
+ "dirs-next",
+ "rustversion",
+ "winapi",
+]
+
 [[package]]
 name = "termcolor"
 version = "1.1.2"
@@ -2149,79 +2223,24 @@ checksum = "360dfd1d6d30e05fda32ace2c8c70e9c0a9da713275777f5a4dbb8a1893930c6"

 [[package]]
 name = "tracing"
-version = "0.1.29"
+version = "0.1.26"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "375a639232caf30edfc78e8d89b2d4c375515393e7af7e16f01cd96917fb2105"
+checksum = "09adeb8c97449311ccd28a427f96fb563e7fd31aabf994189879d9da2394b89d"
 dependencies = [
 "cfg-if 1.0.0",
 "pin-project-lite",
- "tracing-attributes",
 "tracing-core",
 ]

-[[package]]
-name = "tracing-attributes"
-version = "0.1.18"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f4f480b8f81512e825f337ad51e94c1eb5d3bbdf2b363dcd01e2b19a9ffe3f8e"
-dependencies = [
- "proc-macro2",
- "quote",
- "syn",
-]
-
 [[package]]
 name = "tracing-core"
-version = "0.1.21"
+version = "0.1.18"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "1f4ed65637b8390770814083d20756f87bfa2c21bf2f110babdc5438351746e4"
+checksum = "a9ff14f98b1a4b289c6248a023c1c2fa1491062964e9fed67ab29c4e4da4a052"
 dependencies = [
 "lazy_static",
 ]

-[[package]]
-name = "tracing-log"
-version = "0.1.2"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a6923477a48e41c1951f1999ef8bb5a3023eb723ceadafe78ffb65dc366761e3"
-dependencies = [
- "lazy_static",
- "log",
- "tracing-core",
-]
-
-[[package]]
-name = "tracing-serde"
-version = "0.1.2"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "fb65ea441fbb84f9f6748fd496cf7f63ec9af5bca94dd86456978d055e8eb28b"
-dependencies = [
- "serde",
- "tracing-core",
-]
-
-[[package]]
-name = "tracing-subscriber"
-version = "0.2.25"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "0e0d2eaa99c3c2e41547cfa109e910a68ea03823cccad4a0525dcbc9b01e8c71"
-dependencies = [
- "ansi_term 0.12.1",
- "chrono",
- "lazy_static",
- "matchers",
- "regex",
- "serde",
- "serde_json",
- "sharded-slab",
- "smallvec",
- "thread_local",
- "tracing",
- "tracing-core",
- "tracing-log",
- "tracing-serde",
-]
-
 [[package]]
 name = "try-lock"
 version = "0.2.3"
@@ -2326,7 +2345,6 @@ dependencies = [
 "fs2",
 "hex",
 "humantime",
- "hyper",
 "lazy_static",
 "log",
 "pageserver",
@@ -2334,7 +2352,6 @@ dependencies = [
 "postgres-protocol",
 "postgres_ffi",
 "regex",
- "routerify",
 "rust-s3",
 "serde",
 "serde_json",
@@ -2575,6 +2592,7 @@ dependencies = [
 "bincode",
 "byteorder",
 "bytes",
+ "crossbeam-utils",
 "hex",
 "hex-literal",
 "hyper",
@@ -2588,12 +2606,14 @@ dependencies = [
 "rustls-split",
 "serde",
 "serde_json",
+ "slog",
+ "slog-async",
+ "slog-scope",
+ "slog-stdlog",
+ "slog-term",
 "tempfile",
 "thiserror",
 "tokio",
- "tracing",
- "tracing-log",
- "tracing-subscriber",
 "webpki",
 "workspace_hack",
 "zenith_metrics",
--- a/4
+++ b/4
@@ -37,9 +37,7 @@ RUN apt-get update && apt-get -yq install libreadline-dev libseccomp-dev openssl
    mkdir zenith_install

 COPY --from=build /zenith/target/release/pageserver /usr/local/bin
-COPY --from=build /zenith/target/release/safekeeper /usr/local/bin
-# TODO: temporary alias for compatibility, see https://github.com/zenithdb/zenith/pull/740
-RUN ln -s /usr/local/bin/safekeeper /usr/local/bin/wal_acceptor
+COPY --from=build /zenith/target/release/wal_acceptor /usr/local/bin
 COPY --from=build /zenith/target/release/proxy /usr/local/bin
 COPY --from=pg-build /zenith/tmp_install postgres_install
 COPY docker-entrypoint.sh /docker-entrypoint.sh
--- a/Dockerfile.alpine
+++ b/Dockerfile.alpine
@@ -81,9 +81,7 @@ FROM alpine:3.13
 RUN apk add --update openssl build-base libseccomp-dev
 RUN apk --no-cache --update --repository https://dl-cdn.alpinelinux.org/alpine/edge/testing add rocksdb
 COPY --from=build /zenith/target/release/pageserver /usr/local/bin
-COPY --from=build /zenith/target/release/safekeeper /usr/local/bin
-# TODO: temporary alias for compatibility, see https://github.com/zenithdb/zenith/pull/740
-RUN ln -s /usr/local/bin/safekeeper /usr/local/bin/wal_acceptor
+COPY --from=build /zenith/target/release/wal_acceptor /usr/local/bin
 COPY --from=build /zenith/target/release/proxy /usr/local/bin
 COPY --from=pg-build /zenith/tmp_install /usr/local
 COPY docker-entrypoint.sh /docker-entrypoint.sh
--- a/46
+++ b/46
@@ -10,43 +10,32 @@ endif
 # We differentiate between release / debug build types using the BUILD_TYPE
 # environment variable.
 #
-BUILD_TYPE ?= debug
 ifeq ($(BUILD_TYPE),release)
 	PG_CONFIGURE_OPTS = --enable-debug
-	PG_CFLAGS = -O2 -g3 $(CFLAGS)
-	# Unfortunately, `--profile=...` is a nightly feature
-	CARGO_BUILD_FLAGS += --release
-else ifeq ($(BUILD_TYPE),debug)
-	PG_CONFIGURE_OPTS = --enable-debug --enable-cassert --enable-depend
-	PG_CFLAGS = -O0 -g3 $(CFLAGS)
+	PG_CFLAGS = -O2 -g3 ${CFLAGS}
 else
-$(error Bad build type `$(BUILD_TYPE)', see Makefile for options)
+	PG_CONFIGURE_OPTS = --enable-debug --enable-cassert --enable-depend
+	PG_CFLAGS = -O0 -g3 ${CFLAGS}
 endif

-# Choose whether we should be silent or verbose
-CARGO_BUILD_FLAGS += --$(if $(filter s,$(MAKEFLAGS)),quiet,verbose)
-# Fix for a corner case when make doesn't pass a jobserver
-CARGO_BUILD_FLAGS += $(filter -j1,$(MAKEFLAGS))
-
-# This option has a side effect of passing make jobserver to cargo.
-# However, we shouldn't do this if `make -n` (--dry-run) has been asked.
-CARGO_CMD_PREFIX += $(if $(filter n,$(MAKEFLAGS)),,+)
-# Force cargo not to print progress bar
-CARGO_CMD_PREFIX += CARGO_TERM_PROGRESS_WHEN=never CI=1
-
 #
 # Top level Makefile to build Zenith and PostgreSQL
 #
-.PHONY: all
 all: zenith postgres

+# We don't want to run 'cargo build' in parallel with the postgres build,
+# because interleaving cargo build output with postgres build output looks
+# confusing. Also, 'cargo build' is parallel on its own, so it would be too
+# much parallelism. (Recursive invocation of postgres target still gets any
+# '-j' flag from the command line, so 'make -j' is still useful.)
+.NOTPARALLEL:
+
 ### Zenith Rust bits
 #
 # The 'postgres_ffi' depends on the Postgres headers.
 .PHONY: zenith
 zenith: postgres-headers
-	+@echo "Compiling Zenith"
-	$(CARGO_CMD_PREFIX) cargo build $(CARGO_BUILD_FLAGS)
+	cargo build

 ### PostgreSQL parts
 tmp_install/build/config.status:
@@ -68,10 +57,10 @@ postgres-headers: postgres-configure
 	+@echo "Installing PostgreSQL headers"
 	$(MAKE) -C tmp_install/build/src/include MAKELEVEL=0 install

+
 # Compile and install PostgreSQL and contrib/zenith
 .PHONY: postgres
-postgres: postgres-configure \
-		  postgres-headers # to prevent `make install` conflicts with zenith's `postgres-headers`
+postgres: postgres-configure
 	+@echo "Compiling PostgreSQL"
 	$(MAKE) -C tmp_install/build MAKELEVEL=0 install
 	+@echo "Compiling contrib/zenith"
@@ -79,21 +68,18 @@ postgres: postgres-configure \
 	+@echo "Compiling contrib/zenith_test_utils"
 	$(MAKE) -C tmp_install/build/contrib/zenith_test_utils install

-.PHONY: postgres-clean
 postgres-clean:
 	$(MAKE) -C tmp_install/build MAKELEVEL=0 clean

 # This doesn't remove the effects of 'configure'.
-.PHONY: clean
 clean:
-	cd tmp_install/build && $(MAKE) clean
-	$(CARGO_CMD_PREFIX) cargo clean
+	cd tmp_install/build && ${MAKE} clean
+	cargo clean

 # This removes everything
-.PHONY: distclean
 distclean:
 	rm -rf tmp_install
-	$(CARGO_CMD_PREFIX) cargo clean
+	cargo clean

 .PHONY: fmt
 fmt:
--- a/control_plane/src/storage.rs
+++ b/control_plane/src/storage.rs
@@ -85,20 +85,20 @@ impl PageServerNode {
            ),
            env: env.clone(),
            http_client: Client::new(),
-            http_base_url: format!("http://127.0.0.1:{}/v1", env.pageserver_http_port),
+            http_base_url: format!("http://localhost:{}/v1", env.pageserver_http_port),
        }
    }

    fn pageserver_connection_config(password: &str, port: u16) -> Config {
-        format!("postgresql://no_user:{}@127.0.0.1:{}/no_db", password, port)
+        format!("postgresql://no_user:{}@localhost:{}/no_db", password, port)
            .parse()
            .unwrap()
    }

    pub fn init(&self, create_tenant: Option<&str>, enable_auth: bool) -> anyhow::Result<()> {
        let mut cmd = Command::new(self.env.pageserver_bin()?);
-        let listen_pg = format!("127.0.0.1:{}", self.env.pageserver_pg_port);
-        let listen_http = format!("127.0.0.1:{}", self.env.pageserver_http_port);
+        let listen_pg = format!("localhost:{}", self.env.pageserver_pg_port);
+        let listen_http = format!("localhost:{}", self.env.pageserver_http_port);
        let mut args = vec![
            "--init",
            "-D",
@@ -199,45 +199,23 @@ impl PageServerNode {
        bail!("pageserver failed to start in {} seconds", RETRIES);
    }

-    pub fn stop(&self, immediate: bool) -> anyhow::Result<()> {
+    pub fn stop(&self) -> anyhow::Result<()> {
        let pid = read_pidfile(&self.pid_file())?;
        let pid = Pid::from_raw(pid);
-        if immediate {
-            println!("Stop pageserver immediately");
-            if kill(pid, Signal::SIGQUIT).is_err() {
-                bail!("Failed to kill pageserver with pid {}", pid);
-            }
-        } else {
-            println!("Stop pageserver gracefully");
-            if kill(pid, Signal::SIGTERM).is_err() {
-                bail!("Failed to stop pageserver with pid {}", pid);
-            }
+        if kill(pid, Signal::SIGTERM).is_err() {
+            bail!("Failed to kill pageserver with pid {}", pid);
        }

+        // wait for pageserver stop
        let address = connection_address(&self.pg_connection_config);
-
-        // TODO Remove this "timeout" and handle it on caller side instead.
-        // Shutting down may take a long time,
-        // if pageserver checkpoints a lot of data
-        for _ in 0..100 {
-            if let Err(_e) = TcpStream::connect(&address) {
-                println!("Pageserver stopped receiving connections");
-
-                //Now check status
-                match self.check_status() {
-                    Ok(_) => {
-                        println!("Pageserver status is OK. Wait a bit.");
-                        thread::sleep(Duration::from_secs(1));
-                    }
-                    Err(err) => {
-                        println!("Pageserver status is: {}", err);
-                        return Ok(());
-                    }
-                }
-            } else {
-                println!("Pageserver still receives connections");
-                thread::sleep(Duration::from_secs(1));
+        for _ in 0..5 {
+            let stream = TcpStream::connect(&address);
+            thread::sleep(Duration::from_secs(1));
+            if let Err(_e) = stream {
+                println!("Pageserver stopped");
+                return Ok(());
            }
+            println!("Stopping pageserver on {}", address);
        }

        bail!("Failed to stop pageserver with pid {}", pid);
@@ -335,9 +313,8 @@ impl PageServerNode {

 impl Drop for PageServerNode {
    fn drop(&mut self) {
-        // TODO Looks like this flag is never set
        if self.kill_on_exit {
-            let _ = self.stop(true);
+            let _ = self.stop();
        }
    }
 }
--- a/docker-entrypoint.sh
+++ b/docker-entrypoint.sh
@@ -7,7 +7,7 @@ if [ "$1" = 'pageserver' ]; then
        pageserver --init -D /data --postgres-distrib /usr/local
    fi
    echo "Staring pageserver at 0.0.0.0:6400"
-    pageserver -l 0.0.0.0:6400 --listen-http 0.0.0.0:9898 -D /data
+    pageserver -l 0.0.0.0:6400 -D /data
 else
    "$@"
 fi
--- a/docs/docker.md
+++ b/docs/docker.md
@@ -4,7 +4,7 @@

 Currently we build two main images:

- [zenithdb/zenith](https://hub.docker.com/repository/docker/zenithdb/zenith) — image with pre-built `pageserver`, `safekeeper` and `proxy` binaries and all the required runtime dependencies. Built from [/Dockerfile](/Dockerfile).
+- [zenithdb/zenith](https://hub.docker.com/repository/docker/zenithdb/zenith) — image with pre-built `pageserver`, `wal_acceptor` and `proxy` binaries and all the required runtime dependencies. Built from [/Dockerfile](/Dockerfile).
 - [zenithdb/compute-node](https://hub.docker.com/repository/docker/zenithdb/compute-node) — compute node image with pre-built Postgres binaries from [zenithdb/postgres](https://github.com/zenithdb/postgres).

 And two intermediate images used either to reduce build time or to deliver some additional binary tools from other repos:
--- a/docs/multitenancy.md
+++ b/docs/multitenancy.md
@@ -56,4 +56,4 @@ Tenant id is passed to postgres via GUC the same way as the timeline. Tenant id

 ### Safety

-For now particular tenant can only appear on a particular pageserver. Set of safekeepers are also pinned to particular (tenantid, timeline) pair so there can only be one writer for particular (tenantid, timeline).
+For now particular tenant can only appear on a particular pageserver. Set of WAL acceptors are also pinned to particular (tenantid, timeline) pair so there can only be one writer for particular (tenantid, timeline).
--- a/pageserver/Cargo.toml
+++ b/pageserver/Cargo.toml
@@ -17,7 +17,7 @@ lazy_static = "1.4.0"
 log = "0.4.14"
 clap = "2.33.0"
 daemonize = "0.4.1"
-tokio = { version = "1.11", features = ["process", "macros", "fs", "rt"] }
+tokio = { version = "1.11", features = ["process", "macros", "fs"] }
 postgres-types = { git = "https://github.com/zenithdb/rust-postgres.git", rev="9eb0dbfbeb6a6c1b79099b9f7ae4a8c021877858" }
 postgres-protocol = { git = "https://github.com/zenithdb/rust-postgres.git", rev="9eb0dbfbeb6a6c1b79099b9f7ae4a8c021877858" }
 postgres = { git = "https://github.com/zenithdb/rust-postgres.git", rev="9eb0dbfbeb6a6c1b79099b9f7ae4a8c021877858" }
@@ -35,8 +35,6 @@ scopeguard = "1.1.0"
 rust-s3 = { version = "0.27.0-rc4", features = ["no-verify-ssl"] }
 async-trait = "0.1"
 const_format = "0.2.21"
-tracing = "0.1.27"
-signal-hook = {version = "0.3.10", features = ["extended-siginfo"] }

 postgres_ffi = { path = "../postgres_ffi" }
 zenith_metrics = { path = "../zenith_metrics" }
--- a/pageserver/src/basebackup.rs
+++ b/pageserver/src/basebackup.rs
@@ -31,7 +31,7 @@ use zenith_utils::lsn::Lsn;
 pub struct Basebackup<'a> {
    ar: Builder<&'a mut dyn Write>,
    timeline: &'a Arc<dyn Timeline>,
-    pub lsn: Lsn,
+    lsn: Lsn,
    prev_record_lsn: Lsn,
 }

@@ -97,6 +97,7 @@ impl<'a> Basebackup<'a> {
    pub fn send_tarball(&mut self) -> anyhow::Result<()> {
        // Create pgdata subdirs structure
        for dir in pg_constants::PGDATA_SUBDIRS.iter() {
+            info!("send subdir {:?}", *dir);
            let header = new_tar_header_dir(*dir)?;
            self.ar.append(&header, &mut io::empty())?;
        }
--- a/pageserver/src/bin/pageserver.rs
+++ b/pageserver/src/bin/pageserver.rs
@@ -2,6 +2,7 @@
 // Main entry point for the Page Server executable
 //

+use log::*;
 use pageserver::defaults::*;
 use serde::{Deserialize, Serialize};
 use std::{
@@ -11,19 +12,9 @@ use std::{
    str::FromStr,
    thread,
 };
-use tracing::*;
 use zenith_utils::{auth::JwtAuth, logging, postgres_backend::AuthType};

 use anyhow::{bail, ensure, Context, Result};
-use signal_hook::consts::signal::*;
-use signal_hook::consts::TERM_SIGNALS;
-use signal_hook::flag;
-use signal_hook::iterator::exfiltrator::WithOrigin;
-use signal_hook::iterator::SignalsInfo;
-use std::process::exit;
-use std::sync::atomic::AtomicBool;
-use std::sync::Arc;
-
 use clap::{App, Arg, ArgMatches};
 use daemonize::Daemonize;

@@ -37,7 +28,6 @@ use pageserver::{
    RelishStorageKind, S3Config, LOG_FILE_NAME,
 };
 use zenith_utils::http::endpoint;
-use zenith_utils::postgres_backend;

 use const_format::formatcp;

@@ -457,18 +447,7 @@ fn main() -> Result<()> {

 fn start_pageserver(conf: &'static PageServerConf) -> Result<()> {
    // Initialize logger
-    let log_file = logging::init(LOG_FILE_NAME, conf.daemonize)?;
-
-    let term_now = Arc::new(AtomicBool::new(false));
-    for sig in TERM_SIGNALS {
-        // When terminated by a second term signal, exit with exit code 1.
-        // This will do nothing the first time (because term_now is false).
-        flag::register_conditional_shutdown(*sig, 1, Arc::clone(&term_now))?;
-        // But this will "arm" the above for the second time, by setting it to true.
-        // The order of registering these is important, if you put this one first, it will
-        // first arm and then terminate ‒ all in the first round.
-        flag::register(*sig, Arc::clone(&term_now))?;
-    }
+    let (_scope_guard, log_file) = logging::init(LOG_FILE_NAME, conf.daemonize)?;

    // TODO: Check that it looks like a valid repository before going further

@@ -501,7 +480,7 @@ fn start_pageserver(conf: &'static PageServerConf) -> Result<()> {

        match daemonize.start() {
            Ok(_) => info!("Success, daemonized"),
-            Err(err) => error!(%err, "could not daemonize"),
+            Err(e) => error!("could not daemonize: {:#}", e),
        }
    }

@@ -546,42 +525,13 @@ fn start_pageserver(conf: &'static PageServerConf) -> Result<()> {
            page_service::thread_main(conf, auth, pageserver_listener, conf.auth_type)
        })?;

-    for info in SignalsInfo::<WithOrigin>::new(TERM_SIGNALS)?.into_iter() {
-        match info.signal {
-            SIGQUIT => {
-                info!("Got SIGQUIT. Terminate pageserver in immediate shutdown mode");
-                exit(111);
-            }
-            SIGTERM => {
-                info!("Got SIGINT/SIGTERM. Terminate gracefully in fast shutdown mode");
-                // Terminate postgres backends
-                postgres_backend::set_pgbackend_shutdown_requested();
-                // Stop all tenants and flush their data
-                tenant_mgr::shutdown_all_tenants()?;
-                // Wait for pageservice thread to complete the job
-                page_service_thread
-                    .join()
-                    .expect("thread panicked")
-                    .expect("thread exited with an error");
+    join_handles.push(page_service_thread);

-                // Shut down http router
-                endpoint::shutdown();
-
-                // Wait for all threads
-                for handle in join_handles.into_iter() {
-                    handle
-                        .join()
-                        .expect("thread panicked")
-                        .expect("thread exited with an error");
-                }
-                info!("Pageserver shut down successfully completed");
-                exit(0);
-            }
-            _ => {
-                debug!("Unknown signal.");
-            }
-        }
+    for handle in join_handles.into_iter() {
+        handle
+            .join()
+            .expect("thread panicked")
+            .expect("thread exited with an error")
    }
-
    Ok(())
 }
--- a/pageserver/src/branches.rs
+++ b/pageserver/src/branches.rs
@@ -14,12 +14,12 @@ use std::{
    str::FromStr,
    sync::Arc,
 };
-use tracing::*;
+use zenith_utils::zid::{ZTenantId, ZTimelineId};

+use log::*;
 use zenith_utils::crashsafe_dir;
 use zenith_utils::logging;
 use zenith_utils::lsn::Lsn;
-use zenith_utils::zid::{ZTenantId, ZTimelineId};

 use crate::tenant_mgr;
 use crate::walredo::WalRedoManager;
@@ -100,7 +100,7 @@ pub struct PointInTime {
 pub fn init_pageserver(conf: &'static PageServerConf, create_tenant: Option<&str>) -> Result<()> {
    // Initialize logger
    // use true as daemonize parameter because otherwise we pollute zenith cli output with a few pages long output of info messages
-    let _log_file = logging::init(LOG_FILE_NAME, true)?;
+    let (_scope_guard, _log_file) = logging::init(LOG_FILE_NAME, true)?;

    // We don't use the real WAL redo manager, because we don't want to spawn the WAL redo
    // process during repository initialization.
@@ -176,16 +176,13 @@ fn get_lsn_from_controlfile(path: &Path) -> Result<Lsn> {
 // to get bootstrap data for timeline initialization.
 //
 fn run_initdb(conf: &'static PageServerConf, initdbpath: &Path) -> Result<()> {
-    info!("running initdb in {}... ", initdbpath.display());
+    info!("running initdb... ");

    let initdb_path = conf.pg_bin_dir().join("initdb");
    let initdb_output = Command::new(initdb_path)
        .args(&["-D", initdbpath.to_str().unwrap()])
        .args(&["-U", &conf.superuser])
        .arg("--no-instructions")
-        // This is only used for a temporary installation that is deleted shortly after,
-        // so no need to fsync it
-        .arg("--no-sync")
        .env_clear()
        .env("LD_LIBRARY_PATH", conf.pg_lib_dir().to_str().unwrap())
        .env("DYLD_LIBRARY_PATH", conf.pg_lib_dir().to_str().unwrap())
@@ -198,6 +195,7 @@ fn run_initdb(conf: &'static PageServerConf, initdbpath: &Path) -> Result<()> {
            String::from_utf8_lossy(&initdb_output.stderr)
        );
    }
+    info!("initdb succeeded");

    Ok(())
 }
@@ -212,8 +210,6 @@ fn bootstrap_timeline(
    tli: ZTimelineId,
    repo: &dyn Repository,
 ) -> Result<()> {
-    let _enter = info_span!("bootstrapping", timeline = %tli, tenant = %tenantid).entered();
-
    let initdb_path = conf.tenant_path(&tenantid).join("tmp");

    // Init temporarily repo to get bootstrap data
@@ -222,12 +218,14 @@ fn bootstrap_timeline(

    let lsn = get_lsn_from_controlfile(&pgdata_path)?.align();

+    info!("bootstrap_timeline {:?} at lsn {}", pgdata_path, lsn);
+
    // Import the contents of the data directory at the initial checkpoint
    // LSN, and any WAL after that.
    let timeline = repo.create_empty_timeline(tli)?;
    restore_local_repo::import_timeline_from_postgres_datadir(
        &pgdata_path,
-        timeline.writer().as_ref(),
+        timeline.as_ref(),
        lsn,
    )?;
    timeline.checkpoint()?;
--- a/pageserver/src/http/routes.rs
+++ b/pageserver/src/http/routes.rs
@@ -1,3 +1,4 @@
+use std::str::FromStr;
 use std::sync::Arc;

 use anyhow::Result;
@@ -5,7 +6,6 @@ use hyper::header;
 use hyper::StatusCode;
 use hyper::{Body, Request, Response, Uri};
 use routerify::{ext::RequestExt, RouterBuilder};
-use tracing::*;
 use zenith_utils::auth::JwtAuth;
 use zenith_utils::http::endpoint::attach_openapi_ui;
 use zenith_utils::http::endpoint::auth_middleware;
@@ -15,8 +15,6 @@ use zenith_utils::http::{
    endpoint,
    error::HttpErrorBody,
    json::{json_request, json_response},
-    request::get_request_param,
-    request::parse_request_param,
 };

 use super::models::BranchCreateRequest;
@@ -58,6 +56,33 @@ fn get_config(request: &Request<Body>) -> &'static PageServerConf {
    get_state(request).conf
 }

+fn get_request_param<'a>(
+    request: &'a Request<Body>,
+    param_name: &str,
+) -> Result<&'a str, ApiError> {
+    match request.param(param_name) {
+        Some(arg) => Ok(arg),
+        None => {
+            return Err(ApiError::BadRequest(format!(
+                "no {} specified in path param",
+                param_name
+            )))
+        }
+    }
+}
+
+fn parse_request_param<T: FromStr>(
+    request: &Request<Body>,
+    param_name: &str,
+) -> Result<T, ApiError> {
+    match get_request_param(request, param_name)?.parse() {
+        Ok(v) => Ok(v),
+        Err(_) => Err(ApiError::BadRequest(
+            "failed to parse tenant id".to_string(),
+        )),
+    }
+}
+
 // healthcheck handler
 async fn status_handler(_: Request<Body>) -> Result<Response<Body>, ApiError> {
    Ok(Response::builder()
@@ -73,7 +98,6 @@ async fn branch_create_handler(mut request: Request<Body>) -> Result<Response<Bo
    check_permission(&request, Some(request_data.tenant_id))?;

    let response_data = tokio::task::spawn_blocking(move || {
-        let _enter = info_span!("/branch_create", name = %request_data.name, tenant = %request_data.tenant_id, startpoint=%request_data.start_point).entered();
        branches::create_branch(
            get_config(&request),
            &request_data.name,
@@ -92,7 +116,6 @@ async fn branch_list_handler(request: Request<Body>) -> Result<Response<Body>, A
    check_permission(&request, Some(tenantid))?;

    let response_data = tokio::task::spawn_blocking(move || {
-        let _enter = info_span!("branch_list", tenant = %tenantid).entered();
        crate::branches::get_branches(get_config(&request), &tenantid)
    })
    .await
@@ -103,12 +126,11 @@ async fn branch_list_handler(request: Request<Body>) -> Result<Response<Body>, A
 // TODO add to swagger
 async fn branch_detail_handler(request: Request<Body>) -> Result<Response<Body>, ApiError> {
    let tenantid: ZTenantId = parse_request_param(&request, "tenant_id")?;
-    let branch_name: String = get_request_param(&request, "branch_name")?.to_string();
+    let branch_name: &str = get_request_param(&request, "branch_name")?;
    let conf = get_state(&request).conf;
-    let path = conf.branch_path(&branch_name, &tenantid);
+    let path = conf.branch_path(branch_name, &tenantid);

    let response_data = tokio::task::spawn_blocking(move || {
-        let _enter = info_span!("branch_detail", tenant = %tenantid, branch=%branch_name).entered();
        let repo = tenant_mgr::get_repository_for_tenant(tenantid)?;
        BranchInfo::from_path(path, conf, &tenantid, &repo)
    })
@@ -122,13 +144,10 @@ async fn tenant_list_handler(request: Request<Body>) -> Result<Response<Body>, A
    // check for management permission
    check_permission(&request, None)?;

-    let response_data = tokio::task::spawn_blocking(move || {
-        let _enter = info_span!("tenant_list").entered();
-        crate::branches::get_tenants(get_config(&request))
-    })
-    .await
-    .map_err(ApiError::from_err)??;
-
+    let response_data =
+        tokio::task::spawn_blocking(move || crate::branches::get_tenants(get_config(&request)))
+            .await
+            .map_err(ApiError::from_err)??;
    Ok(json_response(StatusCode::OK, response_data)?)
 }

@@ -139,7 +158,6 @@ async fn tenant_create_handler(mut request: Request<Body>) -> Result<Response<Bo
    let request_data: TenantCreateRequest = json_request(&mut request).await?;

    let response_data = tokio::task::spawn_blocking(move || {
-        let _enter = info_span!("tenant_create", tenant = %request_data.tenant_id).entered();
        tenant_mgr::create_repository_for_tenant(get_config(&request), request_data.tenant_id)
    })
    .await
--- a/pageserver/src/layered_repository.rs
+++ b/pageserver/src/layered_repository.rs
@@ -15,9 +15,10 @@ use anyhow::{anyhow, bail, ensure, Context, Result};
 use bookfile::Book;
 use bytes::Bytes;
 use lazy_static::lazy_static;
+use log::*;
 use postgres_ffi::pg_constants::BLCKSZ;
 use serde::{Deserialize, Serialize};
-use tracing::*;
+use zenith_utils::batch_fsync::batch_fsync;

 use std::collections::hash_map::Entry;
 use std::collections::HashMap;
@@ -26,18 +27,16 @@ use std::convert::TryInto;
 use std::fs;
 use std::fs::{File, OpenOptions};
 use std::io::Write;
-use std::ops::{Bound::Included, Deref};
+use std::ops::Bound::Included;
 use std::path::{Path, PathBuf};
 use std::sync::atomic::{AtomicUsize, Ordering};
 use std::sync::{Arc, Mutex, MutexGuard};
-use std::thread::JoinHandle;
 use std::time::{Duration, Instant};

+use crate::layered_repository::inmemory_layer::FreezeLayers;
 use crate::relish::*;
 use crate::relish_storage::schedule_timeline_upload;
-use crate::repository::{GcResult, Repository, Timeline, TimelineWriter, WALRecord};
-use crate::tenant_mgr;
-use crate::walreceiver;
+use crate::repository::{GcResult, Repository, Timeline, WALRecord};
 use crate::walreceiver::IS_WAL_RECEIVER;
 use crate::walredo::WalRedoManager;
 use crate::PageServerConf;
@@ -71,6 +70,8 @@ use storage_layer::{
    Layer, PageReconstructData, PageReconstructResult, SegmentTag, RELISH_SEG_SIZE,
 };

+use self::inmemory_layer::{NonWriteableError, WriteResult};
+
 static ZERO_PAGE: Bytes = Bytes::from_static(&[0u8; 8192]);

 // Timeout when waiting for WAL receiver to catch up to an LSN given in a GetPage@LSN call.
@@ -215,23 +216,6 @@ impl Repository for LayeredRepository {
                self.gc_iteration_internal(target_timelineid, horizon, checkpoint_before_gc)
            })
    }
-
-    // Wait for all threads to complete and persist repository data before pageserver shutdown.
-    fn shutdown(&self) -> Result<()> {
-        trace!("LayeredRepository shutdown for tenant {}", self.tenantid);
-
-        let timelines = self.timelines.lock().unwrap();
-        for (timelineid, timeline) in timelines.iter() {
-            walreceiver::stop_wal_receiver(*timelineid);
-            // Wait for syncing data to disk
-            trace!("repo shutdown. checkpoint timeline {}", timelineid);
-            timeline.checkpoint()?;
-
-            //TODO Wait for walredo process to shutdown too
-        }
-
-        Ok(())
-    }
 }

 /// Private functions
@@ -260,10 +244,6 @@ impl LayeredRepository {
                    None
                };

-                let _enter =
-                    info_span!("loading timeline", timeline = %timelineid, tenant = %self.tenantid)
-                        .entered();
-
                let mut timeline = LayeredTimeline::new(
                    self.conf,
                    metadata,
@@ -315,24 +295,21 @@ impl LayeredRepository {
    ///
    /// Launch the checkpointer thread in given repository.
    ///
-    pub fn launch_checkpointer_thread(
-        conf: &'static PageServerConf,
-        rc: Arc<LayeredRepository>,
-    ) -> JoinHandle<()> {
-        std::thread::Builder::new()
+    pub fn launch_checkpointer_thread(conf: &'static PageServerConf, rc: Arc<LayeredRepository>) {
+        let _thread = std::thread::Builder::new()
            .name("Checkpointer thread".into())
            .spawn(move || {
                // FIXME: relaunch it? Panic is not good.
                rc.checkpoint_loop(conf).expect("Checkpointer thread died");
            })
-            .unwrap()
+            .unwrap();
    }

    ///
    /// Checkpointer thread's main loop
    ///
    fn checkpoint_loop(&self, conf: &'static PageServerConf) -> Result<()> {
-        while !tenant_mgr::shutdown_requested() {
+        loop {
            std::thread::sleep(conf.checkpoint_period);
            info!("checkpointer thread for tenant {} waking up", self.tenantid);

@@ -340,60 +317,44 @@ impl LayeredRepository {
            // bytes of WAL since last checkpoint.
            {
                let timelines = self.timelines.lock().unwrap();
-                for (timelineid, timeline) in timelines.iter() {
-                    let _entered =
-                        info_span!("checkpoint", timeline = %timelineid, tenant = %self.tenantid)
-                            .entered();
-
+                for (_timelineid, timeline) in timelines.iter() {
                    STORAGE_TIME
                        .with_label_values(&["checkpoint_timed"])
                        .observe_closure_duration(|| {
-                            timeline.checkpoint_internal(conf.checkpoint_distance, false)
+                            timeline.checkpoint_internal(conf.checkpoint_distance)
                        })?
                }
                // release lock on 'timelines'
            }
        }
-        trace!("Checkpointer thread shut down");
-        Ok(())
    }

    ///
    /// Launch the GC thread in given repository.
    ///
-    pub fn launch_gc_thread(
-        conf: &'static PageServerConf,
-        rc: Arc<LayeredRepository>,
-    ) -> JoinHandle<()> {
-        std::thread::Builder::new()
+    pub fn launch_gc_thread(conf: &'static PageServerConf, rc: Arc<LayeredRepository>) {
+        let _thread = std::thread::Builder::new()
            .name("GC thread".into())
            .spawn(move || {
                // FIXME: relaunch it? Panic is not good.
                rc.gc_loop(conf).expect("GC thread died");
            })
-            .unwrap()
+            .unwrap();
    }

    ///
    /// GC thread's main loop
    ///
    fn gc_loop(&self, conf: &'static PageServerConf) -> Result<()> {
-        while !tenant_mgr::shutdown_requested() {
+        loop {
+            std::thread::sleep(conf.gc_period);
+            info!("gc thread for tenant {} waking up", self.tenantid);
+
            // Garbage collect old files that are not needed for PITR anymore
            if conf.gc_horizon > 0 {
                self.gc_iteration(None, conf.gc_horizon, false).unwrap();
            }
-
-            // TODO Write it in more adequate way using
-            // condvar.wait_timeout() or something
-            let mut sleep_time = conf.gc_period.as_secs();
-            while sleep_time > 0 && !tenant_mgr::shutdown_requested() {
-                sleep_time -= 1;
-                std::thread::sleep(Duration::from_secs(1));
-            }
-            info!("gc thread for tenant {} waking up", self.tenantid);
        }
-        Ok(())
    }

    /// Save timeline metadata to file
@@ -404,7 +365,6 @@ impl LayeredRepository {
        data: &TimelineMetadata,
        first_save: bool,
    ) -> Result<()> {
-        let _enter = info_span!("saving metadata").entered();
        let path = metadata_path(conf, timelineid, tenantid);
        // use OpenOptions to ensure file presence is consistent with first_save
        let mut file = OpenOptions::new()
@@ -412,6 +372,8 @@ impl LayeredRepository {
            .create_new(first_save)
            .open(&path)?;

+        info!("saving metadata {}", path.display());
+
        let mut metadata_bytes = TimelineMetadata::ser(data)?;

        assert!(metadata_bytes.len() <= METADATA_MAX_DATA_SIZE);
@@ -540,10 +502,6 @@ impl LayeredRepository {
        // Ok, we now know all the branch points.
        // Perform GC for each timeline.
        for timelineid in timelineids {
-            if tenant_mgr::shutdown_requested() {
-                return Ok(totals);
-            }
-
            // We have already loaded all timelines above
            // so this operation is just a quick map lookup.
            let timeline = self.get_timeline_locked(timelineid, &mut *timelines)?;
@@ -667,13 +625,6 @@ pub struct LayeredTimeline {

    /// If `true`, will backup its timeline files to remote storage after freezing.
    upload_relishes: bool,
-
-    /// Ensures layers aren't frozen by checkpointer between
-    /// [`LayeredTimeline::get_layer_for_write`] and layer reads.
-    /// Locked automatically by [`LayeredTimelineWriter`] and checkpointer.
-    /// Must always be acquired before the layer map/individual layer lock
-    /// to avoid deadlock.
-    write_lock: Mutex<()>,
 }

 /// Public interface functions
@@ -736,7 +687,13 @@ impl Timeline for LayeredTimeline {
            let segsize;
            if let Some((layer, lsn)) = self.get_layer_for_read(seg, lsn)? {
                segsize = layer.get_seg_size(lsn)?;
-                trace!("get_seg_size: {} at {} -> {}", seg, lsn, segsize);
+                trace!(
+                    "get_seg_size: {} at {}/{} -> {}",
+                    seg,
+                    self.timelineid,
+                    lsn,
+                    segsize
+                );
            } else {
                if segno == 0 {
                    return Ok(None);
@@ -838,13 +795,138 @@ impl Timeline for LayeredTimeline {
                result.insert(new_relish);
                trace!("List object {}", new_relish);
            } else {
-                trace!("Filtered out dropped object {}", new_relish);
+                trace!("Filter out droped object {}", new_relish);
            }
        }

        Ok(result)
    }

+    fn put_wal_record(&self, rel: RelishTag, blknum: u32, rec: WALRecord) -> Result<()> {
+        if !rel.is_blocky() && blknum != 0 {
+            bail!(
+                "invalid request for block {} for non-blocky relish {}",
+                blknum,
+                rel
+            );
+        }
+        ensure!(rec.lsn.is_aligned(), "unaligned record LSN");
+
+        let seg = SegmentTag::from_blknum(rel, blknum);
+        let delta_size = self.perform_write_op(seg, rec.lsn, |layer| {
+            layer.put_wal_record(blknum, rec.clone())
+        })?;
+        self.increase_current_logical_size(delta_size * BLCKSZ as u32);
+        Ok(())
+    }
+
+    fn put_truncation(&self, rel: RelishTag, lsn: Lsn, relsize: u32) -> anyhow::Result<()> {
+        if !rel.is_blocky() {
+            bail!("invalid truncation for non-blocky relish {}", rel);
+        }
+        ensure!(lsn.is_aligned(), "unaligned record LSN");
+
+        debug!("put_truncation: {} to {} blocks at {}", rel, relsize, lsn);
+
+        let oldsize = self
+            .get_relish_size(rel, self.get_last_record_lsn())?
+            .ok_or_else(|| {
+                anyhow!(
+                    "attempted to truncate non-existent relish {} at {}",
+                    rel,
+                    lsn
+                )
+            })?;
+
+        if oldsize <= relsize {
+            return Ok(());
+        }
+        let old_last_seg = (oldsize - 1) / RELISH_SEG_SIZE;
+
+        let last_remain_seg = if relsize == 0 {
+            0
+        } else {
+            (relsize - 1) / RELISH_SEG_SIZE
+        };
+
+        // Drop segments beyond the last remaining segment.
+        for remove_segno in (last_remain_seg + 1)..=old_last_seg {
+            let seg = SegmentTag {
+                rel,
+                segno: remove_segno,
+            };
+            self.perform_write_op(seg, lsn, |layer| layer.drop_segment(lsn))?;
+        }
+
+        // Truncate the last remaining segment to the specified size
+        if relsize == 0 || relsize % RELISH_SEG_SIZE != 0 {
+            let seg = SegmentTag {
+                rel,
+                segno: last_remain_seg,
+            };
+            self.perform_write_op(seg, lsn, |layer| {
+                layer.put_truncation(lsn, relsize % RELISH_SEG_SIZE)
+            })?;
+        }
+        self.decrease_current_logical_size((oldsize - relsize) * BLCKSZ as u32);
+        Ok(())
+    }
+
+    fn drop_relish(&self, rel: RelishTag, lsn: Lsn) -> Result<()> {
+        trace!("drop_segment: {} at {}", rel, lsn);
+
+        if rel.is_blocky() {
+            if let Some(oldsize) = self.get_relish_size(rel, self.get_last_record_lsn())? {
+                let old_last_seg = if oldsize == 0 {
+                    0
+                } else {
+                    (oldsize - 1) / RELISH_SEG_SIZE
+                };
+
+                // Drop all segments of the relish
+                for remove_segno in 0..=old_last_seg {
+                    let seg = SegmentTag {
+                        rel,
+                        segno: remove_segno,
+                    };
+                    self.perform_write_op(seg, lsn, |layer| layer.drop_segment(lsn))?;
+                }
+                self.decrease_current_logical_size(oldsize * BLCKSZ as u32);
+            } else {
+                warn!(
+                    "drop_segment called on non-existent relish {} at {}",
+                    rel, lsn
+                );
+            }
+        } else {
+            // TODO handle TwoPhase relishes
+            let seg = SegmentTag::from_blknum(rel, 0);
+            self.perform_write_op(seg, lsn, |layer| layer.drop_segment(lsn))?;
+        }
+
+        Ok(())
+    }
+
+    fn put_page_image(&self, rel: RelishTag, blknum: u32, lsn: Lsn, img: Bytes) -> Result<()> {
+        if !rel.is_blocky() && blknum != 0 {
+            bail!(
+                "invalid request for block {} for non-blocky relish {}",
+                blknum,
+                rel
+            );
+        }
+        ensure!(lsn.is_aligned(), "unaligned record LSN");
+
+        let seg = SegmentTag::from_blknum(rel, blknum);
+
+        let delta_size = self.perform_write_op(seg, lsn, |layer| {
+            layer.put_page_image(blknum, lsn, img.clone())
+        })?;
+
+        self.increase_current_logical_size(delta_size * BLCKSZ as u32);
+        Ok(())
+    }
+
    /// Public entry point for checkpoint(). All the logic is in the private
    /// checkpoint_internal function, this public facade just wraps it for
    /// metrics collection.
@@ -852,7 +934,16 @@ impl Timeline for LayeredTimeline {
        STORAGE_TIME
            .with_label_values(&["checkpoint_force"])
            //pass checkpoint_distance=0 to force checkpoint
-            .observe_closure_duration(|| self.checkpoint_internal(0, true))
+            .observe_closure_duration(|| self.checkpoint_internal(0))
+    }
+
+    ///
+    /// Remember the (end of) last valid WAL record remembered in the timeline.
+    ///
+    fn advance_last_record_lsn(&self, new_lsn: Lsn) {
+        assert!(new_lsn.is_aligned());
+
+        self.last_record_lsn.advance(new_lsn);
    }

    fn get_last_record_lsn(&self) -> Lsn {
@@ -882,8 +973,6 @@ impl Timeline for LayeredTimeline {
    fn get_current_logical_size_non_incremental(&self, lsn: Lsn) -> Result<usize> {
        let mut total_blocks: usize = 0;

-        let _enter = info_span!("calc logical size", %lsn).entered();
-
        // list of all relations in this timeline, including ancestor timelines
        let all_rels = self.list_rels(0, 0, lsn)?;

@@ -905,13 +994,6 @@ impl Timeline for LayeredTimeline {

        Ok(total_blocks * BLCKSZ as usize)
    }
-
-    fn writer<'a>(&'a self) -> Box<dyn TimelineWriter + 'a> {
-        Box::new(LayeredTimelineWriter {
-            tl: self,
-            _write_guard: self.write_lock.lock().unwrap(),
-        })
-    }
 }

 impl LayeredTimeline {
@@ -952,8 +1034,6 @@ impl LayeredTimeline {
            current_logical_size: AtomicUsize::new(current_logical_size),
            current_logical_size_gauge,
            upload_relishes,
-
-            write_lock: Mutex::new(()),
        };
        Ok(timeline)
    }
@@ -963,8 +1043,11 @@ impl LayeredTimeline {
    /// Returns all timeline-related files that were found and loaded.
    ///
    fn load_layer_map(&self, disk_consistent_lsn: Lsn) -> anyhow::Result<Vec<PathBuf>> {
+        info!(
+            "loading layer map for timeline {} into memory",
+            self.timelineid
+        );
        let mut layers = self.layers.lock().unwrap();
-        let mut num_layers = 0;
        let (imgfilenames, deltafilenames) =
            filename::list_files(self.conf, self.timelineid, self.tenantid)?;

@@ -984,10 +1067,14 @@ impl LayeredTimeline {

            let layer = ImageLayer::new(self.conf, self.timelineid, self.tenantid, filename);

-            trace!("found layer {}", layer.filename().display());
+            info!(
+                "found layer {} {} on timeline {}",
+                layer.get_seg_tag(),
+                layer.get_start_lsn(),
+                self.timelineid
+            );
            local_layers.push(layer.path());
            layers.insert_historic(Arc::new(layer));
-            num_layers += 1;
        }

        for filename in &deltafilenames {
@@ -1004,12 +1091,14 @@ impl LayeredTimeline {

            let layer = DeltaLayer::new(self.conf, self.timelineid, self.tenantid, filename);

-            trace!("found layer {}", layer.filename().display());
+            info!(
+                "found layer {} on timeline {}",
+                layer.filename().display(),
+                self.timelineid,
+            );
            local_layers.push(layer.path());
            layers.insert_historic(Arc::new(layer));
-            num_layers += 1;
        }
-        info!("loaded layer map with {} layers", num_layers);

        Ok(local_layers)
    }
@@ -1060,7 +1149,12 @@ impl LayeredTimeline {
        lsn: Lsn,
        self_layers: &MutexGuard<LayerMap>,
    ) -> Result<Option<(Arc<dyn Layer>, Lsn)>> {
-        trace!("get_layer_for_read called for {} at {}", seg, lsn);
+        trace!(
+            "get_layer_for_read called for {} at {}/{}",
+            seg,
+            self.timelineid,
+            lsn
+        );

        // If you requested a page at an older LSN, before the branch point, dig into
        // the right ancestor timeline. This can only happen if you launch a read-only
@@ -1178,15 +1272,17 @@ impl LayeredTimeline {
                // First modification on this timeline
                start_lsn = self.ancestor_lsn + 1;
                trace!(
-                    "creating layer for write for {} at branch point {}",
+                    "creating layer for write for {} at branch point {}/{}",
                    seg,
+                    self.timelineid,
                    start_lsn
                );
            } else {
                start_lsn = prev_layer.get_end_lsn();
                trace!(
-                    "creating layer for write for {} after previous layer {}",
+                    "creating layer for write for {} after previous layer {}/{}",
                    seg,
+                    self.timelineid,
                    start_lsn
                );
            }
@@ -1227,20 +1323,29 @@ impl LayeredTimeline {
    /// Flush to disk all data that was written with the put_* functions
    ///
    /// NOTE: This has nothing to do with checkpoint in PostgreSQL.
-    fn checkpoint_internal(&self, checkpoint_distance: u64, forced: bool) -> Result<()> {
-        let mut write_guard = self.write_lock.lock().unwrap();
+    fn checkpoint_internal(&self, checkpoint_distance: u64) -> Result<()> {
+        // Grab lock on the layer map.
+        //
+        // TODO: We hold it locked throughout the checkpoint operation. That's bad,
+        // the checkpointing could take many seconds, and any incoming get_page_at_lsn()
+        // requests will block.
        let mut layers = self.layers.lock().unwrap();

        // Bump the generation number in the layer map, so that we can distinguish
        // entries inserted after the checkpoint started
        let current_generation = layers.increment_generation();

+        // Read 'last_record_lsn'. That becomes the cutoff LSN for frozen layers.
        let RecordLsn {
            last: last_record_lsn,
            prev: prev_record_lsn,
        } = self.last_record_lsn.load();

-        trace!("checkpoint starting at {}", last_record_lsn);
+        trace!(
+            "checkpointing timeline {} at {}",
+            self.timelineid,
+            last_record_lsn
+        );

        // Take the in-memory layer with the oldest WAL record. If it's older
        // than the threshold, write it out to disk as a new image and delta file.
@@ -1258,10 +1363,6 @@ impl LayeredTimeline {
        while let Some((oldest_layer, oldest_generation)) = layers.peek_oldest_open() {
            let oldest_pending_lsn = oldest_layer.get_oldest_pending_lsn();

-            if tenant_mgr::shutdown_requested() && !forced {
-                return Ok(());
-            }
-
            // Does this layer need freezing?
            //
            // Write out all in-memory layers that contain WAL older than CHECKPOINT_DISTANCE.
@@ -1284,24 +1385,32 @@ impl LayeredTimeline {
                break;
            }

-            // Mark the layer as no longer accepting writes and record the end_lsn.
-            // This happens in-place, no new layers are created now.
-            // We call `get_last_record_lsn` again, which may be different from the
-            // original load, as we may have released the write lock since then.
-            oldest_layer.freeze(self.get_last_record_lsn());
+            // Freeze the layer.
+            //
+            // This is a two-step process. First, we "freeze" the in-memory
+            // layer, to close it for new writes, and replace the original
+            // layer with the new frozen in-memory layer (and possibly a new
+            // open layer to hold changes newer than the cutoff.) Then we write
+            // the frozen layer to disk, and replace the in-memory frozen layer
+            // with the new on-disk layers.
+            let FreezeLayers {
+                frozen,
+                open: maybe_new_open,
+            } = oldest_layer.freeze(last_record_lsn)?;

-            // The layer is no longer open, update the layer map to reflect this.
-            // We will replace it with on-disk historics below.
+            // replace this layer with the new layers that 'freeze' returned
            layers.pop_oldest_open();
-            layers.insert_historic(oldest_layer.clone());
+            if let Some(new_open) = maybe_new_open.clone() {
+                layers.insert_open(new_open);
+            }
+
+            // We temporarily insert InMemory layer into historic list here.
+            // TODO: check that all possible concurrent users of 'historic' treat it right
+            layers.insert_historic(frozen.clone());

            // Write the now-frozen layer to disk. That could take a while, so release the lock while do it
            drop(layers);
-            drop(write_guard);
-
-            let new_historics = oldest_layer.write_to_disk(self)?;
-
-            write_guard = self.write_lock.lock().unwrap();
+            let new_historics = frozen.write_to_disk(self)?;
            layers = self.layers.lock().unwrap();

            if !new_historics.is_empty() {
@@ -1309,7 +1418,7 @@ impl LayeredTimeline {
            }

            // Finally, replace the frozen in-memory layer with the new on-disk layers
-            layers.remove_historic(oldest_layer);
+            layers.remove_historic(frozen.clone());

            // Add the historics to the LayerMap
            for delta_layer in new_historics.delta_layers {
@@ -1330,14 +1439,13 @@ impl LayeredTimeline {
        }

        drop(layers);
-        drop(write_guard);

        if created_historics {
            // We must fsync the timeline dir to ensure the directory entries for
            // new layer files are durable
-            let timeline_dir =
-                File::open(self.conf.timeline_path(&self.timelineid, &self.tenantid))?;
-            timeline_dir.sync_all()?;
+            layer_uploads.push(self.conf.timeline_path(&self.timelineid, &self.tenantid));
+            batch_fsync(&layer_uploads)?;
+            layer_uploads.pop().unwrap();
        }

        // Save the metadata, with updated 'disk_consistent_lsn', to a
@@ -1413,11 +1521,11 @@ impl LayeredTimeline {
        let now = Instant::now();
        let mut result: GcResult = Default::default();

-        let _enter = info_span!("garbage collection", timeline = %self.timelineid, tenant = %self.tenantid, cutoff = %cutoff).entered();
-
-        info!("GC starting");
-
-        debug!("retain_lsns: {:?}", retain_lsns);
+        info!(
+            "running GC on timeline {}, cutoff {}",
+            self.timelineid, cutoff
+        );
+        info!("retain_lsns:  {:?}", retain_lsns);

        let mut layers_to_remove: Vec<Arc<dyn Layer>> = Vec::new();

@@ -1679,9 +1787,10 @@ impl LayeredTimeline {
        if data.records.is_empty() {
            if let Some(img) = &data.page_img {
                trace!(
-                    "found page image for blk {} in {} at {}, no WAL redo required",
+                    "found page image for blk {} in {} at {}/{}, no WAL redo required",
                    blknum,
                    rel,
+                    self.timelineid,
                    request_lsn
                );
                Ok(img.clone())
@@ -1695,7 +1804,7 @@ impl LayeredTimeline {
            //
            // If we don't have a base image, then the oldest WAL record better initialize
            // the page
-            if data.page_img.is_none() && !data.records.first().unwrap().1.will_init {
+            if data.page_img.is_none() && !data.records.first().unwrap().will_init {
                // FIXME: this ought to be an error?
                warn!(
                    "Base image for page {}/{} at {} not found, but got {} WAL records",
@@ -1707,9 +1816,9 @@ impl LayeredTimeline {
                Ok(ZERO_PAGE.clone())
            } else {
                if data.page_img.is_some() {
-                    trace!("found {} WAL records and a base image for blk {} in {} at {}, performing WAL redo", data.records.len(), blknum, rel, request_lsn);
+                    trace!("found {} WAL records and a base image for blk {} in {} at {}/{}, performing WAL redo", data.records.len(), blknum, rel, self.timelineid, request_lsn);
                } else {
-                    trace!("found {} WAL records that will init the page for blk {} in {} at {}, performing WAL redo", data.records.len(), blknum, rel, request_lsn);
+                    trace!("found {} WAL records that will init the page for blk {} in {} at {}/{}, performing WAL redo", data.records.len(), blknum, rel, self.timelineid, request_lsn);
                }
                let img = self.walredo_mgr.request_redo(
                    rel,
@@ -1757,163 +1866,36 @@ impl LayeredTimeline {
        self.current_logical_size_gauge
            .set(val as i64 - diff as i64);
    }
-}

-struct LayeredTimelineWriter<'a> {
-    tl: &'a LayeredTimeline,
-    _write_guard: MutexGuard<'a, ()>,
-}
-
-impl Deref for LayeredTimelineWriter<'_> {
-    type Target = dyn Timeline;
-
-    fn deref(&self) -> &Self::Target {
-        self.tl
-    }
-}
-
-impl<'a> TimelineWriter for LayeredTimelineWriter<'a> {
-    fn put_wal_record(&self, lsn: Lsn, rel: RelishTag, blknum: u32, rec: WALRecord) -> Result<()> {
-        if !rel.is_blocky() && blknum != 0 {
-            bail!(
-                "invalid request for block {} for non-blocky relish {}",
-                blknum,
-                rel
-            );
-        }
-        ensure!(lsn.is_aligned(), "unaligned record LSN");
-
-        let seg = SegmentTag::from_blknum(rel, blknum);
-        let layer = self.tl.get_layer_for_write(seg, lsn)?;
-        let delta_size = layer.put_wal_record(lsn, blknum, rec);
-        self.tl
-            .increase_current_logical_size(delta_size * BLCKSZ as u32);
-        Ok(())
-    }
-
-    fn put_page_image(&self, rel: RelishTag, blknum: u32, lsn: Lsn, img: Bytes) -> Result<()> {
-        if !rel.is_blocky() && blknum != 0 {
-            bail!(
-                "invalid request for block {} for non-blocky relish {}",
-                blknum,
-                rel
-            );
-        }
-        ensure!(lsn.is_aligned(), "unaligned record LSN");
-
-        let seg = SegmentTag::from_blknum(rel, blknum);
-
-        let layer = self.tl.get_layer_for_write(seg, lsn)?;
-        let delta_size = layer.put_page_image(blknum, lsn, img);
-
-        self.tl
-            .increase_current_logical_size(delta_size * BLCKSZ as u32);
-        Ok(())
-    }
-
-    fn put_truncation(&self, rel: RelishTag, lsn: Lsn, relsize: u32) -> Result<()> {
-        if !rel.is_blocky() {
-            bail!("invalid truncation for non-blocky relish {}", rel);
-        }
-        ensure!(lsn.is_aligned(), "unaligned record LSN");
-
-        debug!("put_truncation: {} to {} blocks at {}", rel, relsize, lsn);
-
-        let oldsize = self
-            .tl
-            .get_relish_size(rel, self.tl.get_last_record_lsn())?
-            .ok_or_else(|| {
-                anyhow!(
-                    "attempted to truncate non-existent relish {} at {}",
-                    rel,
-                    lsn
-                )
-            })?;
-
-        if oldsize <= relsize {
-            return Ok(());
-        }
-        let old_last_seg = (oldsize - 1) / RELISH_SEG_SIZE;
-
-        let last_remain_seg = if relsize == 0 {
-            0
-        } else {
-            (relsize - 1) / RELISH_SEG_SIZE
-        };
-
-        // Drop segments beyond the last remaining segment.
-        for remove_segno in (last_remain_seg + 1)..=old_last_seg {
-            let seg = SegmentTag {
-                rel,
-                segno: remove_segno,
-            };
-
-            let layer = self.tl.get_layer_for_write(seg, lsn)?;
-            layer.drop_segment(lsn);
-        }
-
-        // Truncate the last remaining segment to the specified size
-        if relsize == 0 || relsize % RELISH_SEG_SIZE != 0 {
-            let seg = SegmentTag {
-                rel,
-                segno: last_remain_seg,
-            };
-            let layer = self.tl.get_layer_for_write(seg, lsn)?;
-            layer.put_truncation(lsn, relsize % RELISH_SEG_SIZE)
-        }
-        self.tl
-            .decrease_current_logical_size((oldsize - relsize) * BLCKSZ as u32);
-        Ok(())
-    }
-
-    fn drop_relish(&self, rel: RelishTag, lsn: Lsn) -> Result<()> {
-        trace!("drop_segment: {} at {}", rel, lsn);
-
-        if rel.is_blocky() {
-            if let Some(oldsize) = self
-                .tl
-                .get_relish_size(rel, self.tl.get_last_record_lsn())?
-            {
-                let old_last_seg = if oldsize == 0 {
-                    0
-                } else {
-                    (oldsize - 1) / RELISH_SEG_SIZE
-                };
-
-                // Drop all segments of the relish
-                for remove_segno in 0..=old_last_seg {
-                    let seg = SegmentTag {
-                        rel,
-                        segno: remove_segno,
-                    };
-                    let layer = self.tl.get_layer_for_write(seg, lsn)?;
-                    layer.drop_segment(lsn);
-                }
-                self.tl
-                    .decrease_current_logical_size(oldsize * BLCKSZ as u32);
-            } else {
-                warn!(
-                    "drop_segment called on non-existent relish {} at {}",
-                    rel, lsn
-                );
+    /// If a layer is in the process of being replaced in [`LayerMap`], write
+    /// operations will fail with [`NonWriteableError`]. This may happen due to
+    /// a race: the checkpointer thread freezes a layer just after
+    /// [`Self::get_layer_for_write`] returned it. To handle this error, we try
+    /// again getting the layer and attempt the write.
+    fn perform_write_op<R>(
+        &self,
+        seg: SegmentTag,
+        lsn: Lsn,
+        write_op: impl Fn(&Arc<InMemoryLayer>) -> WriteResult<R>,
+    ) -> anyhow::Result<R> {
+        let mut layer = self.get_layer_for_write(seg, lsn)?;
+        loop {
+            match write_op(&layer) {
+                Ok(r) => return Ok(r),
+                Err(NonWriteableError {}) => {}
            }
-        } else {
-            // TODO handle TwoPhase relishes
-            let seg = SegmentTag::from_blknum(rel, 0);
-            let layer = self.tl.get_layer_for_write(seg, lsn)?;
-            layer.drop_segment(lsn);
+
+            info!(
+                "attempted to write to non-writeable layer, retrying {} {}",
+                seg, lsn
+            );
+
+            // layer was non-writeable, try again
+            let new_layer = self.get_layer_for_write(seg, lsn)?;
+            // the new layer does not have to be writeable, but it should at least be different
+            assert!(!Arc::ptr_eq(&layer, &new_layer));
+            layer = new_layer;
        }
-
-        Ok(())
-    }
-
-    ///
-    /// Remember the (end of) last valid WAL record remembered in the timeline.
-    ///
-    fn advance_last_record_lsn(&self, new_lsn: Lsn) {
-        assert!(new_lsn.is_aligned());
-
-        self.tl.last_record_lsn.advance(new_lsn);
    }
 }

--- a/pageserver/src/layered_repository/delta_layer.rs
+++ b/pageserver/src/layered_repository/delta_layer.rs
@@ -198,7 +198,7 @@ impl Layer for DeltaLayer {
                .slice_range((Included(&minkey), Included(&maxkey)))
                .iter()
                .rev();
-            for ((_blknum, pv_lsn), blob_range) in iter {
+            for ((_blknum, _lsn), blob_range) in iter {
                let pv = PageVersion::des(&read_blob(&page_version_reader, blob_range)?)?;

                if let Some(img) = pv.page_image {
@@ -208,7 +208,7 @@ impl Layer for DeltaLayer {
                    break;
                } else if let Some(rec) = pv.record {
                    let will_init = rec.will_init;
-                    reconstruct_data.records.push((*pv_lsn, rec));
+                    reconstruct_data.records.push(rec);
                    if will_init {
                        // This WAL record initializes the page, so no need to go further back
                        need_image = false;
@@ -431,8 +431,7 @@ impl DeltaLayer {
        let book = chapter.close()?;

        // This flushes the underlying 'buf_writer'.
-        let writer = book.close()?;
-        writer.get_ref().sync_all()?;
+        book.close()?;

        trace!("saved {}", &path.display());

--- a/pageserver/src/layered_repository/image_layer.rs
+++ b/pageserver/src/layered_repository/image_layer.rs
@@ -315,8 +315,7 @@ impl ImageLayer {
        let book = chapter.close()?;

        // This flushes the underlying 'buf_writer'.
-        let writer = book.close()?;
-        writer.get_ref().sync_all()?;
+        book.close()?;

        trace!("saved {}", path.display());

--- a/pageserver/src/layered_repository/inmemory_layer.rs
+++ b/pageserver/src/layered_repository/inmemory_layer.rs
@@ -15,10 +15,12 @@ use crate::{ZTenantId, ZTimelineId};
 use anyhow::{bail, ensure, Result};
 use bytes::Bytes;
 use log::*;
+use std::cmp::min;
 use std::path::PathBuf;
 use std::sync::{Arc, RwLock};
 use zenith_utils::vec_map::VecMap;

+use zenith_utils::accum::Accum;
 use zenith_utils::lsn::Lsn;

 use super::page_versions::PageVersions;
@@ -35,6 +37,9 @@ pub struct InMemoryLayer {
    ///
    start_lsn: Lsn,

+    /// Frozen in-memory layers have an inclusive end LSN.
+    end_lsn: Option<Lsn>,
+
    /// LSN of the oldest page version stored in this layer
    oldest_pending_lsn: Lsn,

@@ -47,13 +52,8 @@ pub struct InMemoryLayer {
 }

 pub struct InMemoryLayerInner {
-    /// Frozen in-memory layers have an exclusive end LSN.
-    /// Writes are only allowed when this is None
-    end_lsn: Option<Lsn>,
-
    /// If this relation was dropped, remember when that happened.
-    /// The drop LSN is recorded in [`end_lsn`].
-    dropped: bool,
+    drop_lsn: Option<Lsn>,

    ///
    /// All versions of all pages in the layer are are kept here.
@@ -69,11 +69,19 @@ pub struct InMemoryLayerInner {
    /// a non-blocky rel, 'segsizes' is not used and is always empty.
    ///
    segsizes: VecMap<Lsn, u32>,
+
+    /// Writes are only allowed when true.
+    /// Set to false when this layer is in the process of being replaced.
+    writeable: bool,
 }

 impl InMemoryLayerInner {
-    fn assert_writeable(&self) {
-        assert!(self.end_lsn.is_none());
+    fn check_writeable(&self) -> WriteResult<()> {
+        if self.writeable {
+            Ok(())
+        } else {
+            Err(NonWriteableError)
+        }
    }

    fn get_seg_size(&self, lsn: Lsn) -> u32 {
@@ -96,17 +104,20 @@ impl Layer for InMemoryLayer {
        let inner = self.inner.read().unwrap();

        let end_lsn;
-        if let Some(drop_lsn) = inner.end_lsn {
+        let dropped;
+        if let Some(drop_lsn) = inner.drop_lsn {
            end_lsn = drop_lsn;
+            dropped = true;
        } else {
            end_lsn = Lsn(u64::MAX);
+            dropped = false;
        }

        let delta_filename = DeltaFileName {
            seg: self.seg,
            start_lsn: self.start_lsn,
            end_lsn,
-            dropped: inner.dropped,
+            dropped,
        }
        .to_string();

@@ -126,10 +137,14 @@ impl Layer for InMemoryLayer {
    }

    fn get_end_lsn(&self) -> Lsn {
+        if let Some(end_lsn) = self.end_lsn {
+            return Lsn(end_lsn.0 + 1);
+        }
+
        let inner = self.inner.read().unwrap();

-        if let Some(end_lsn) = inner.end_lsn {
-            end_lsn
+        if let Some(drop_lsn) = inner.drop_lsn {
+            drop_lsn
        } else {
            Lsn(u64::MAX)
        }
@@ -137,7 +152,7 @@ impl Layer for InMemoryLayer {

    fn is_dropped(&self) -> bool {
        let inner = self.inner.read().unwrap();
-        inner.dropped
+        inner.drop_lsn.is_some()
    }

    /// Look up given page in the cache.
@@ -160,13 +175,13 @@ impl Layer for InMemoryLayer {
                .get_block_lsn_range(blknum, ..=lsn)
                .iter()
                .rev();
-            for (entry_lsn, entry) in iter {
+            for (_entry_lsn, entry) in iter {
                if let Some(img) = &entry.page_image {
                    reconstruct_data.page_img = Some(img.clone());
                    need_image = false;
                    break;
                } else if let Some(rec) = &entry.record {
-                    reconstruct_data.records.push((*entry_lsn, rec.clone()));
+                    reconstruct_data.records.push(rec.clone());
                    if rec.will_init {
                        // This WAL record initializes the page, so no need to go further back
                        need_image = false;
@@ -215,8 +230,8 @@ impl Layer for InMemoryLayer {
        assert!(lsn >= self.start_lsn);

        // Is the requested LSN after the segment was dropped?
-        if let Some(end_lsn) = inner.end_lsn {
-            if lsn >= end_lsn {
+        if let Some(drop_lsn) = inner.drop_lsn {
+            if lsn >= drop_lsn {
                return Ok(false);
            }
        }
@@ -247,14 +262,14 @@ impl Layer for InMemoryLayer {
        let inner = self.inner.read().unwrap();

        let end_str = inner
-            .end_lsn
+            .drop_lsn
            .as_ref()
-            .map(Lsn::to_string)
+            .map(|drop_lsn| drop_lsn.to_string())
            .unwrap_or_default();

        println!(
-            "----- in-memory layer for tli {} seg {} {}-{} {} ----",
-            self.timelineid, self.seg, self.start_lsn, end_str, inner.dropped,
+            "----- in-memory layer for tli {} seg {} {}-{} ----",
+            self.timelineid, self.seg, self.start_lsn, end_str
        );

        for (k, v) in inner.segsizes.as_slice() {
@@ -275,6 +290,21 @@ impl Layer for InMemoryLayer {
    }
 }

+/// Write failed because the layer is in process of being replaced.
+/// See [`LayeredTimeline::perform_write_op`] for how to handle this error.
+#[derive(Debug)]
+pub struct NonWriteableError;
+
+pub type WriteResult<T> = std::result::Result<T, NonWriteableError>;
+
+/// Helper struct to cleanup `InMemoryLayer::freeze` return signature.
+pub struct FreezeLayers {
+    /// Replacement layer for the layer which freeze was called on.
+    pub frozen: Arc<InMemoryLayer>,
+    /// New open layer containing leftover data.
+    pub open: Option<Arc<InMemoryLayer>>,
+}
+
 /// A result of an inmemory layer data being written to disk.
 pub struct LayersOnDisk {
    pub delta_layers: Vec<DeltaLayer>,
@@ -288,6 +318,10 @@ impl LayersOnDisk {
 }

 impl InMemoryLayer {
+    fn assert_not_frozen(&self) {
+        assert!(self.end_lsn.is_none());
+    }
+
    /// Return the oldest page version that's stored in this layer
    pub fn get_oldest_pending_lsn(&self) -> Lsn {
        self.oldest_pending_lsn
@@ -323,13 +357,14 @@ impl InMemoryLayer {
            tenantid,
            seg,
            start_lsn,
+            end_lsn: None,
            oldest_pending_lsn,
            incremental: false,
            inner: RwLock::new(InMemoryLayerInner {
-                end_lsn: None,
-                dropped: false,
+                drop_lsn: None,
                page_versions: PageVersions::default(),
                segsizes,
+                writeable: true,
            }),
        })
    }
@@ -337,10 +372,10 @@ impl InMemoryLayer {
    // Write operations

    /// Remember new page version, as a WAL record over previous version
-    pub fn put_wal_record(&self, lsn: Lsn, blknum: u32, rec: WALRecord) -> u32 {
+    pub fn put_wal_record(&self, blknum: u32, rec: WALRecord) -> WriteResult<u32> {
        self.put_page_version(
            blknum,
-            lsn,
+            rec.lsn,
            PageVersion {
                page_image: None,
                record: Some(rec),
@@ -349,7 +384,7 @@ impl InMemoryLayer {
    }

    /// Remember new page version, as a full page image
-    pub fn put_page_image(&self, blknum: u32, lsn: Lsn, img: Bytes) -> u32 {
+    pub fn put_page_image(&self, blknum: u32, lsn: Lsn, img: Bytes) -> WriteResult<u32> {
        self.put_page_version(
            blknum,
            lsn,
@@ -362,7 +397,8 @@ impl InMemoryLayer {

    /// Common subroutine of the public put_wal_record() and put_page_image() functions.
    /// Adds the page version to the in-memory tree
-    pub fn put_page_version(&self, blknum: u32, lsn: Lsn, pv: PageVersion) -> u32 {
+    pub fn put_page_version(&self, blknum: u32, lsn: Lsn, pv: PageVersion) -> WriteResult<u32> {
+        self.assert_not_frozen();
        assert!(self.seg.blknum_in_seg(blknum));

        trace!(
@@ -374,7 +410,7 @@ impl InMemoryLayer {
        );
        let mut inner = self.inner.write().unwrap();

-        inner.assert_writeable();
+        inner.check_writeable()?;

        let old = inner.page_versions.append_or_update_last(blknum, lsn, pv);

@@ -435,22 +471,22 @@ impl InMemoryLayer {
                }

                inner.segsizes.append_or_update_last(lsn, newsize).unwrap();
-                return newsize - oldsize;
+                return Ok(newsize - oldsize);
            }
        }
-
-        0
+        Ok(0)
    }

    /// Remember that the relation was truncated at given LSN
-    pub fn put_truncation(&self, lsn: Lsn, segsize: u32) {
+    pub fn put_truncation(&self, lsn: Lsn, segsize: u32) -> WriteResult<()> {
        assert!(
            self.seg.rel.is_blocky(),
            "put_truncation() called on a non-blocky rel"
        );
+        self.assert_not_frozen();

        let mut inner = self.inner.write().unwrap();
-        inner.assert_writeable();
+        inner.check_writeable()?;

        // check that this we truncate to a smaller size than segment was before the truncation
        let oldsize = inner.get_seg_size(lsn);
@@ -462,19 +498,25 @@ impl InMemoryLayer {
            // We already had an entry for this LSN. That's odd..
            warn!("Inserting truncation, but had an entry for the LSN already");
        }
+
+        Ok(())
    }

    /// Remember that the segment was dropped at given LSN
-    pub fn drop_segment(&self, lsn: Lsn) {
+    pub fn drop_segment(&self, lsn: Lsn) -> WriteResult<()> {
+        self.assert_not_frozen();
+
        let mut inner = self.inner.write().unwrap();

-        assert!(inner.end_lsn.is_none());
-        assert!(!inner.dropped);
-        inner.dropped = true;
-        assert!(self.start_lsn < lsn);
-        inner.end_lsn = Some(lsn);
+        inner.check_writeable()?;
+
+        assert!(inner.drop_lsn.is_none());
+        inner.drop_lsn = Some(lsn);
+        inner.writeable = false;

        trace!("dropped segment {} at {}", self.seg, lsn);
+
+        Ok(())
    }

    ///
@@ -514,43 +556,117 @@ impl InMemoryLayer {
            tenantid,
            seg,
            start_lsn,
+            end_lsn: None,
            oldest_pending_lsn,
            incremental: true,
            inner: RwLock::new(InMemoryLayerInner {
-                end_lsn: None,
-                dropped: false,
+                drop_lsn: None,
                page_versions: PageVersions::default(),
                segsizes,
+                writeable: true,
            }),
        })
    }

    pub fn is_writeable(&self) -> bool {
        let inner = self.inner.read().unwrap();
-        inner.end_lsn.is_none()
+        inner.writeable
    }

-    /// Make the layer non-writeable. Only call once.
-    /// Records the end_lsn for non-dropped layers.
-    /// `end_lsn` is inclusive
-    pub fn freeze(&self, end_lsn: Lsn) {
-        let mut inner = self.inner.write().unwrap();
+    /// Splits `self` into two InMemoryLayers: `frozen` and `open`.
+    /// All data up to and including `cutoff_lsn`
+    /// is copied to `frozen`, while the remaining data is copied to `open`.
+    /// After completion, self is non-writeable, but not frozen.
+    pub fn freeze(self: Arc<Self>, cutoff_lsn: Lsn) -> Result<FreezeLayers> {
+        info!(
+            "freezing in memory layer {} on timeline {} at {} (oldest {})",
+            self.filename().display(),
+            self.timelineid,
+            cutoff_lsn,
+            self.oldest_pending_lsn
+        );

-        if inner.end_lsn.is_some() {
-            assert!(inner.dropped);
-        } else {
-            assert!(!inner.dropped);
-            assert!(self.start_lsn < end_lsn + 1);
-            inner.end_lsn = Some(Lsn(end_lsn.0 + 1));
+        self.assert_not_frozen();

-            if let Some((lsn, _)) = inner.segsizes.as_slice().last() {
-                assert!(lsn <= &end_lsn, "{:?} {:?}", lsn, end_lsn);
-            }
+        let self_ref = self.clone();
+        let mut inner = self_ref.inner.write().unwrap();
+        // Dropped layers don't need any special freeze actions,
+        // they are marked as non-writeable at drop and just
+        // written out to disk by checkpointer.
+        if inner.drop_lsn.is_some() {
+            assert!(!inner.writeable);
+            info!(
+                "freezing in memory layer for {} on timeline {} is dropped at {}",
+                self.seg,
+                self.timelineid,
+                inner.drop_lsn.unwrap()
+            );

-            for (_blk, lsn, _pv) in inner.page_versions.ordered_page_version_iter(None) {
-                assert!(lsn <= end_lsn);
-            }
+            // There should be no newer layer that refers this non-writeable layer,
+            // because layer that is created after dropped one represents a new rel.
+            return Ok(FreezeLayers {
+                frozen: self,
+                open: None,
+            });
        }
+        assert!(inner.writeable);
+        inner.writeable = false;
+
+        // Divide all the page versions into old and new
+        // at the 'cutoff_lsn' point.
+        let mut after_oldest_lsn: Accum<Lsn> = Accum(None);
+
+        let cutoff_lsn_exclusive = Lsn(cutoff_lsn.0 + 1);
+
+        let (before_segsizes, mut after_segsizes) = inner.segsizes.split_at(&cutoff_lsn_exclusive);
+        if let Some((lsn, _size)) = after_segsizes.as_slice().first() {
+            after_oldest_lsn.accum(min, *lsn);
+        }
+
+        let (before_page_versions, after_page_versions) = inner
+            .page_versions
+            .split_at(cutoff_lsn_exclusive, &mut after_oldest_lsn);
+
+        let frozen = Arc::new(InMemoryLayer {
+            conf: self.conf,
+            tenantid: self.tenantid,
+            timelineid: self.timelineid,
+            seg: self.seg,
+            start_lsn: self.start_lsn,
+            end_lsn: Some(cutoff_lsn),
+            oldest_pending_lsn: self.start_lsn,
+            incremental: self.incremental,
+            inner: RwLock::new(InMemoryLayerInner {
+                drop_lsn: inner.drop_lsn,
+                page_versions: before_page_versions,
+                segsizes: before_segsizes,
+                writeable: false,
+            }),
+        });
+
+        let open = if !after_segsizes.is_empty() || !after_page_versions.is_empty() {
+            let mut new_open = Self::create_successor_layer(
+                self.conf,
+                frozen.clone(),
+                self.timelineid,
+                self.tenantid,
+                cutoff_lsn + 1,
+                after_oldest_lsn.0.unwrap(),
+            )?;
+
+            let new_inner = new_open.inner.get_mut().unwrap();
+            // Ensure page_versions doesn't contain anything
+            // so we can just replace it
+            assert!(new_inner.page_versions.is_empty());
+            new_inner.page_versions = after_page_versions;
+            new_inner.segsizes.extend(&mut after_segsizes).unwrap();
+
+            Some(Arc::new(new_open))
+        } else {
+            None
+        };
+
+        Ok(FreezeLayers { frozen, open })
    }

    /// Write the this frozen in-memory layer to disk.
@@ -563,8 +679,9 @@ impl InMemoryLayer {
    /// end LSN are the same.)
    pub fn write_to_disk(&self, timeline: &LayeredTimeline) -> Result<LayersOnDisk> {
        trace!(
-            "write_to_disk {} get_end_lsn is {}",
+            "write_to_disk {} end_lsn is {} get_end_lsn is {}",
            self.filename().display(),
+            self.end_lsn.unwrap_or(Lsn(0)),
            self.get_end_lsn()
        );

@@ -578,16 +695,16 @@ impl InMemoryLayer {
        // would have to wait until we release it. That race condition is very
        // rare though, so we just accept the potential latency hit for now.
        let inner = self.inner.read().unwrap();
-        let end_lsn_exclusive = inner.end_lsn.unwrap();
+        assert!(!inner.writeable);

-        if inner.dropped {
+        if let Some(drop_lsn) = inner.drop_lsn {
            let delta_layer = DeltaLayer::create(
                self.conf,
                self.timelineid,
                self.tenantid,
                self.seg,
                self.start_lsn,
-                end_lsn_exclusive,
+                drop_lsn,
                true,
                inner.page_versions.ordered_page_version_iter(None),
                inner.segsizes.clone(),
@@ -596,7 +713,7 @@ impl InMemoryLayer {
                "freeze: created delta layer for dropped segment {} {}-{}",
                self.seg,
                self.start_lsn,
-                end_lsn_exclusive
+                drop_lsn
            );
            return Ok(LayersOnDisk {
                delta_layers: vec![delta_layer],
@@ -604,19 +721,14 @@ impl InMemoryLayer {
            });
        }

-        // Since `end_lsn` is inclusive, subtract 1.
-        // We want to make an ImageLayer for the last included LSN,
-        // so the DeltaLayer should exlcude that LSN.
-        let end_lsn_inclusive = Lsn(end_lsn_exclusive.0 - 1);
+        let end_lsn = self.end_lsn.unwrap();

-        let mut page_versions = inner
-            .page_versions
-            .ordered_page_version_iter(Some(end_lsn_inclusive));
+        let mut before_page_versions = inner.page_versions.ordered_page_version_iter(Some(end_lsn));

        let mut delta_layers = Vec::new();
+        if self.start_lsn != end_lsn {
+            let (before_segsizes, _after_segsizes) = inner.segsizes.split_at(&Lsn(end_lsn.0 + 1));

-        if self.start_lsn != end_lsn_inclusive {
-            let (segsizes, _) = inner.segsizes.split_at(&end_lsn_exclusive);
            // Write the page versions before the cutoff to disk.
            let delta_layer = DeltaLayer::create(
                self.conf,
@@ -624,32 +736,27 @@ impl InMemoryLayer {
                self.tenantid,
                self.seg,
                self.start_lsn,
-                end_lsn_inclusive,
+                end_lsn,
                false,
-                page_versions,
-                segsizes,
+                before_page_versions,
+                before_segsizes,
            )?;
            delta_layers.push(delta_layer);
            trace!(
                "freeze: created delta layer {} {}-{}",
                self.seg,
                self.start_lsn,
-                end_lsn_inclusive
+                end_lsn
            );
        } else {
-            assert!(page_versions.next().is_none());
+            assert!(before_page_versions.next().is_none());
        }

        drop(inner);

        // Write a new base image layer at the cutoff point
-        let image_layer =
-            ImageLayer::create_from_src(self.conf, timeline, self, end_lsn_inclusive)?;
-        trace!(
-            "freeze: created image layer {} at {}",
-            self.seg,
-            end_lsn_inclusive
-        );
+        let image_layer = ImageLayer::create_from_src(self.conf, timeline, self, end_lsn)?;
+        trace!("freeze: created image layer {} at {}", self.seg, end_lsn);

        Ok(LayersOnDisk {
            delta_layers,
--- a/pageserver/src/layered_repository/page_versions.rs
+++ b/pageserver/src/layered_repository/page_versions.rs
@@ -1,6 +1,6 @@
 use std::{collections::HashMap, ops::RangeBounds, slice};

-use zenith_utils::{lsn::Lsn, vec_map::VecMap};
+use zenith_utils::{accum::Accum, lsn::Lsn, vec_map::VecMap};

 use super::storage_layer::PageVersion;

@@ -10,6 +10,10 @@ const EMPTY_SLICE: &[(Lsn, PageVersion)] = &[];
 pub struct PageVersions(HashMap<u32, VecMap<Lsn, PageVersion>>);

 impl PageVersions {
+    pub fn is_empty(&self) -> bool {
+        self.0.is_empty()
+    }
+
    pub fn append_or_update_last(
        &mut self,
        blknum: u32,
@@ -40,6 +44,34 @@ impl PageVersions {
            .unwrap_or(EMPTY_SLICE)
    }

+    /// Split the page version map into two.
+    ///
+    /// Left contains everything up to and not including [`cutoff_lsn`].
+    /// Right contains [`cutoff_lsn`] and everything after.
+    pub fn split_at(&self, cutoff_lsn: Lsn, after_oldest_lsn: &mut Accum<Lsn>) -> (Self, Self) {
+        let mut before_blocks = HashMap::new();
+        let mut after_blocks = HashMap::new();
+
+        for (blknum, vec_map) in self.0.iter() {
+            let (before_versions, after_versions) = vec_map.split_at(&cutoff_lsn);
+
+            if !before_versions.is_empty() {
+                let old = before_blocks.insert(*blknum, before_versions);
+                assert!(old.is_none());
+            }
+
+            if !after_versions.is_empty() {
+                let (first_lsn, _first_pv) = &after_versions.as_slice()[0];
+                after_oldest_lsn.accum(std::cmp::min, *first_lsn);
+
+                let old = after_blocks.insert(*blknum, after_versions);
+                assert!(old.is_none());
+            }
+        }
+
+        (Self(before_blocks), Self(after_blocks))
+    }
+
    /// Iterate through [`PageVersion`]s in (block, lsn) order.
    /// If a [`cutoff_lsn`] is set, only show versions with `lsn < cutoff_lsn`
    pub fn ordered_page_version_iter(&self, cutoff_lsn: Option<Lsn>) -> OrderedPageVersionIter<'_> {
--- a/pageserver/src/layered_repository/storage_layer.rs
+++ b/pageserver/src/layered_repository/storage_layer.rs
@@ -78,7 +78,7 @@ pub struct PageVersion {
 /// 'records' contains the records to apply over the base image.
 ///
 pub struct PageReconstructData {
-    pub records: Vec<(Lsn, WALRecord)>,
+    pub records: Vec<WALRecord>,
    pub page_img: Option<Bytes>,
 }

--- a/pageserver/src/page_service.rs
+++ b/pageserver/src/page_service.rs
@@ -13,6 +13,7 @@
 use anyhow::{anyhow, bail, ensure, Result};
 use bytes::{Buf, BufMut, Bytes, BytesMut};
 use lazy_static::lazy_static;
+use log::*;
 use regex::Regex;
 use std::net::TcpListener;
 use std::str;
@@ -20,12 +21,10 @@ use std::str::FromStr;
 use std::sync::Arc;
 use std::thread;
 use std::{io, net::TcpStream};
-use tracing::*;
 use zenith_metrics::{register_histogram_vec, HistogramVec};
 use zenith_utils::auth::{self, JwtAuth};
 use zenith_utils::auth::{Claims, Scope};
 use zenith_utils::lsn::Lsn;
-use zenith_utils::postgres_backend::is_socket_read_timed_out;
 use zenith_utils::postgres_backend::PostgresBackend;
 use zenith_utils::postgres_backend::{self, AuthType};
 use zenith_utils::pq_proto::{
@@ -188,32 +187,17 @@ pub fn thread_main(
    listener: TcpListener,
    auth_type: AuthType,
 ) -> anyhow::Result<()> {
-    let mut join_handles = Vec::new();
-
-    while !tenant_mgr::shutdown_requested() {
+    loop {
        let (socket, peer_addr) = listener.accept()?;
        debug!("accepted connection from {}", peer_addr);
        socket.set_nodelay(true).unwrap();
        let local_auth = auth.clone();
-
-        let handle = thread::Builder::new()
-            .name("serving Page Service thread".into())
-            .spawn(move || {
-                if let Err(err) = page_service_conn_main(conf, local_auth, socket, auth_type) {
-                    error!(%err, "page server thread exited with error");
-                }
-            })
-            .unwrap();
-
-        join_handles.push(handle);
+        thread::spawn(move || {
+            if let Err(err) = page_service_conn_main(conf, local_auth, socket, auth_type) {
+                error!("page server thread exiting with error: {:#}", err);
+            }
+        });
    }
-
-    debug!("page_service loop terminated. wait for connections to cancel");
-    for handle in join_handles.into_iter() {
-        handle.join().unwrap();
-    }
-
-    Ok(())
 }

 fn page_service_conn_main(
@@ -232,7 +216,7 @@ fn page_service_conn_main(
    }

    let mut conn_handler = PageServerHandler::new(conf, auth);
-    let pgbackend = PostgresBackend::new(socket, auth_type, None, true)?;
+    let pgbackend = PostgresBackend::new(socket, auth_type, None)?;
    pgbackend.run(&mut conn_handler)
 }

@@ -276,66 +260,50 @@ impl PageServerHandler {
        timelineid: ZTimelineId,
        tenantid: ZTenantId,
    ) -> anyhow::Result<()> {
-        let _enter = info_span!("pagestream", timeline = %timelineid, tenant = %tenantid).entered();
-
        // Check that the timeline exists
        let timeline = tenant_mgr::get_timeline_for_tenant(tenantid, timelineid)?;

        /* switch client to COPYBOTH */
        pgb.write_message(&BeMessage::CopyBothResponse)?;

-        while !tenant_mgr::shutdown_requested() {
-            match pgb.read_message() {
-                Ok(message) => {
-                    if let Some(message) = message {
-                        trace!("query: {:?}", message);
+        while let Some(message) = pgb.read_message()? {
+            trace!("query({:?}): {:?}", timelineid, message);

-                        let copy_data_bytes = match message {
-                            FeMessage::CopyData(bytes) => bytes,
-                            _ => continue,
-                        };
+            let copy_data_bytes = match message {
+                FeMessage::CopyData(bytes) => bytes,
+                _ => continue,
+            };

-                        let zenith_fe_msg = PagestreamFeMessage::parse(copy_data_bytes)?;
+            let zenith_fe_msg = PagestreamFeMessage::parse(copy_data_bytes)?;

-                        let response = match zenith_fe_msg {
-                            PagestreamFeMessage::Exists(req) => SMGR_QUERY_TIME
-                                .with_label_values(&["get_rel_exists"])
-                                .observe_closure_duration(|| {
-                                    self.handle_get_rel_exists_request(&*timeline, &req)
-                                }),
-                            PagestreamFeMessage::Nblocks(req) => SMGR_QUERY_TIME
-                                .with_label_values(&["get_rel_size"])
-                                .observe_closure_duration(|| {
-                                    self.handle_get_nblocks_request(&*timeline, &req)
-                                }),
-                            PagestreamFeMessage::GetPage(req) => SMGR_QUERY_TIME
-                                .with_label_values(&["get_page_at_lsn"])
-                                .observe_closure_duration(|| {
-                                    self.handle_get_page_at_lsn_request(&*timeline, &req)
-                                }),
-                        };
+            let response = match zenith_fe_msg {
+                PagestreamFeMessage::Exists(req) => SMGR_QUERY_TIME
+                    .with_label_values(&["get_rel_exists"])
+                    .observe_closure_duration(|| {
+                        self.handle_get_rel_exists_request(&*timeline, &req)
+                    }),
+                PagestreamFeMessage::Nblocks(req) => SMGR_QUERY_TIME
+                    .with_label_values(&["get_rel_size"])
+                    .observe_closure_duration(|| self.handle_get_nblocks_request(&*timeline, &req)),
+                PagestreamFeMessage::GetPage(req) => SMGR_QUERY_TIME
+                    .with_label_values(&["get_page_at_lsn"])
+                    .observe_closure_duration(|| {
+                        self.handle_get_page_at_lsn_request(&*timeline, &req)
+                    }),
+            };

-                        let response = response.unwrap_or_else(|e| {
-                            // print the all details to the log with {:#}, but for the client the
-                            // error message is enough
-                            error!("error reading relation or page version: {:#}", e);
-                            PagestreamBeMessage::Error(PagestreamErrorResponse {
-                                message: e.to_string(),
-                            })
-                        });
+            let response = response.unwrap_or_else(|e| {
+                // print the all details to the log with {:#}, but for the client the
+                // error message is enough
+                error!("error reading relation or page version: {:#}", e);
+                PagestreamBeMessage::Error(PagestreamErrorResponse {
+                    message: e.to_string(),
+                })
+            });

-                        pgb.write_message(&BeMessage::CopyData(&response.serialize()))?;
-                    } else {
-                        break;
-                    }
-                }
-                Err(e) => {
-                    if !is_socket_read_timed_out(&e) {
-                        return Err(e);
-                    }
-                }
-            }
+            pgb.write_message(&BeMessage::CopyData(&response.serialize()))?;
        }
+
        Ok(())
    }

@@ -395,8 +363,6 @@ impl PageServerHandler {
        timeline: &dyn Timeline,
        req: &PagestreamExistsRequest,
    ) -> Result<PagestreamBeMessage> {
-        let _enter = info_span!("get_rel_exists", rel = %req.rel, req_lsn = %req.lsn).entered();
-
        let tag = RelishTag::Relation(req.rel);
        let lsn = Self::wait_or_get_last_lsn(timeline, req.lsn, req.latest)?;

@@ -412,7 +378,6 @@ impl PageServerHandler {
        timeline: &dyn Timeline,
        req: &PagestreamNblocksRequest,
    ) -> Result<PagestreamBeMessage> {
-        let _enter = info_span!("get_nblocks", rel = %req.rel, req_lsn = %req.lsn).entered();
        let tag = RelishTag::Relation(req.rel);
        let lsn = Self::wait_or_get_last_lsn(timeline, req.lsn, req.latest)?;

@@ -432,8 +397,6 @@ impl PageServerHandler {
        timeline: &dyn Timeline,
        req: &PagestreamGetPageRequest,
    ) -> Result<PagestreamBeMessage> {
-        let _enter = info_span!("get_page", rel = %req.rel, blkno = &req.blkno, req_lsn = %req.lsn)
-            .entered();
        let tag = RelishTag::Relation(req.rel);
        let lsn = Self::wait_or_get_last_lsn(timeline, req.lsn, req.latest)?;

@@ -451,20 +414,17 @@ impl PageServerHandler {
        lsn: Option<Lsn>,
        tenantid: ZTenantId,
    ) -> anyhow::Result<()> {
-        let span = info_span!("basebackup", timeline = %timelineid, tenant = %tenantid, lsn = field::Empty);
-        let _enter = span.enter();
-
        // check that the timeline exists
        let timeline = tenant_mgr::get_timeline_for_tenant(tenantid, timelineid)?;

-        // switch client to COPYOUT
+        /* switch client to COPYOUT */
        pgb.write_message(&BeMessage::CopyOutResponse)?;
+        info!("sent CopyOut");

        /* Send a tarball of the latest layer on the timeline */
        {
            let mut writer = CopyDataSink { pgb };
            let mut basebackup = basebackup::Basebackup::new(&mut writer, &timeline, lsn)?;
-            span.record("lsn", &basebackup.lsn.to_string().as_str());
            basebackup.send_tarball()?;
        }
        pgb.write_message(&BeMessage::CopyDone)?;
@@ -569,6 +529,11 @@ impl postgres_backend::Handler for PageServerHandler {
                None
            };

+            info!(
+                "got basebackup command. tenantid=\"{}\" timelineid=\"{}\" lsn=\"{:#?}\"",
+                tenantid, timelineid, lsn
+            );
+
            // Check that the timeline exists
            self.handle_basebackup_request(pgb, timelineid, lsn, tenantid)?;
            pgb.write_message_noflush(&BeMessage::CommandComplete(b"SELECT 1"))?;
@@ -586,9 +551,6 @@ impl postgres_backend::Handler for PageServerHandler {

            self.check_permission(Some(tenantid))?;

-            let _enter =
-                info_span!("callmemaybe", timeline = %timelineid, tenant = %tenantid).entered();
-
            // Check that the timeline exists
            tenant_mgr::get_timeline_for_tenant(tenantid, timelineid)?;

@@ -611,9 +573,6 @@ impl postgres_backend::Handler for PageServerHandler {

            self.check_permission(Some(tenantid))?;

-            let _enter =
-                info_span!("branch_create", name = %branchname, tenant = %tenantid).entered();
-
            let branch =
                branches::create_branch(self.conf, &branchname, &startpoint_str, &tenantid)?;
            let branch = serde_json::to_vec(&branch)?;
--- a/pageserver/src/relish_storage.rs
+++ b/pageserver/src/relish_storage.rs
@@ -55,21 +55,15 @@ pub trait RelishStorage: Send + Sync {

    async fn list_relishes(&self) -> anyhow::Result<Vec<Self::RelishStoragePath>>;

-    async fn download_relish<W: 'static + std::io::Write + Send>(
+    async fn download_relish(
        &self,
        from: &Self::RelishStoragePath,
-        // rust_s3 `get_object_stream` method requires `std::io::BufWriter` for some reason, not the async counterpart
-        // that forces us to consume and return the writer to satisfy the blocking operation async wrapper requirements
-        to: std::io::BufWriter<W>,
-    ) -> anyhow::Result<std::io::BufWriter<W>>;
+        to: &Path,
+    ) -> anyhow::Result<()>;

    async fn delete_relish(&self, path: &Self::RelishStoragePath) -> anyhow::Result<()>;

-    async fn upload_relish<R: tokio::io::AsyncRead + std::marker::Unpin + Send>(
-        &self,
-        from: &mut tokio::io::BufReader<R>,
-        to: &Self::RelishStoragePath,
-    ) -> anyhow::Result<()>;
+    async fn upload_relish(&self, from: &Path, to: &Self::RelishStoragePath) -> anyhow::Result<()>;
 }

 fn strip_workspace_prefix<'a>(
--- a/pageserver/src/relish_storage/local_fs.rs
+++ b/pageserver/src/relish_storage/local_fs.rs
@@ -9,13 +9,11 @@

 use std::{
    future::Future,
-    io::Write,
    path::{Path, PathBuf},
    pin::Pin,
 };

 use anyhow::{bail, Context};
-use tokio::{fs, io};

 use super::{strip_workspace_prefix, RelishStorage};

@@ -66,33 +64,16 @@ impl RelishStorage for LocalFs {
        Ok(get_all_files(&self.root).await?.into_iter().collect())
    }

-    async fn download_relish<W: 'static + std::io::Write + Send>(
+    async fn download_relish(
        &self,
        from: &Self::RelishStoragePath,
-        mut to: std::io::BufWriter<W>,
-    ) -> anyhow::Result<std::io::BufWriter<W>> {
+        to: &Path,
+    ) -> anyhow::Result<()> {
        let file_path = self.resolve_in_storage(from)?;
        if file_path.exists() && file_path.is_file() {
-            let updated_buffer = tokio::task::spawn_blocking(move || {
-                let mut source = std::io::BufReader::new(
-                    std::fs::OpenOptions::new()
-                        .read(true)
-                        .open(&file_path)
-                        .with_context(|| {
-                            format!(
-                                "Failed to open source file '{}' to use in the download",
-                                file_path.display()
-                            )
-                        })?,
-                );
-                std::io::copy(&mut source, &mut to)
-                    .context("Failed to download the relish file")?;
-                to.flush().context("Failed to flush the download buffer")?;
-                Ok::<_, anyhow::Error>(to)
-            })
-            .await
-            .context("Failed to spawn a blocking task")??;
-            Ok(updated_buffer)
+            create_target_directory(to).await?;
+            tokio::fs::copy(file_path, to).await?;
+            Ok(())
        } else {
            bail!(
                "File '{}' either does not exist or is not a file",
@@ -113,30 +94,18 @@ impl RelishStorage for LocalFs {
        }
    }

-    async fn upload_relish<R: io::AsyncRead + std::marker::Unpin + Send>(
-        &self,
-        from: &mut io::BufReader<R>,
-        to: &Self::RelishStoragePath,
-    ) -> anyhow::Result<()> {
+    async fn upload_relish(&self, from: &Path, to: &Self::RelishStoragePath) -> anyhow::Result<()> {
        let target_file_path = self.resolve_in_storage(to)?;
        create_target_directory(&target_file_path).await?;
-        let mut destination = io::BufWriter::new(
-            fs::OpenOptions::new()
-                .write(true)
-                .create(true)
-                .open(&target_file_path)
-                .await
-                .with_context(|| {
-                    format!(
-                        "Failed to open target fs destination at '{}'",
-                        target_file_path.display()
-                    )
-                })?,
-        );

-        io::copy_buf(from, &mut destination)
+        tokio::fs::copy(&from, &target_file_path)
            .await
-            .context("Failed to upload relish to local storage")?;
+            .with_context(|| {
+                format!(
+                    "Failed to upload relish '{}' to local storage",
+                    from.display(),
+                )
+            })?;
        Ok(())
    }
 }
--- a/pageserver/src/relish_storage/rust_s3.rs
+++ b/pageserver/src/relish_storage/rust_s3.rs
@@ -1,6 +1,5 @@
 //! A wrapper around AWS S3 client library `rust_s3` to be used a relish storage.

-use std::io::Write;
 use std::path::Path;

 use anyhow::Context;
@@ -84,14 +83,18 @@ impl RelishStorage for RustS3 {
            .collect())
    }

-    async fn download_relish<W: 'static + std::io::Write + Send>(
+    async fn download_relish(
        &self,
        from: &Self::RelishStoragePath,
-        mut to: std::io::BufWriter<W>,
-    ) -> anyhow::Result<std::io::BufWriter<W>> {
+        to: &Path,
+    ) -> anyhow::Result<()> {
+        let mut target_file = std::fs::OpenOptions::new()
+            .write(true)
+            .open(to)
+            .with_context(|| format!("Failed to open target s3 destination at {}", to.display()))?;
        let code = self
            .bucket
-            .get_object_stream(from.key(), &mut to)
+            .get_object_stream(from.key(), &mut target_file)
            .await
            .with_context(|| format!("Failed to download s3 object with key {}", from.key()))?;
        if code != 200 {
@@ -100,12 +103,7 @@ impl RelishStorage for RustS3 {
                code
            ))
        } else {
-            tokio::task::spawn_blocking(move || {
-                to.flush().context("Failed to fluch the downoad buffer")?;
-                Ok::<_, anyhow::Error>(to)
-            })
-            .await
-            .context("Failed to joim the download buffer flush task")?
+            Ok(())
        }
    }

@@ -126,14 +124,12 @@ impl RelishStorage for RustS3 {
        }
    }

-    async fn upload_relish<R: tokio::io::AsyncRead + std::marker::Unpin + Send>(
-        &self,
-        from: &mut tokio::io::BufReader<R>,
-        to: &Self::RelishStoragePath,
-    ) -> anyhow::Result<()> {
+    async fn upload_relish(&self, from: &Path, to: &Self::RelishStoragePath) -> anyhow::Result<()> {
+        let mut local_file = tokio::fs::OpenOptions::new().read(true).open(from).await?;
+
        let code = self
            .bucket
-            .put_object_stream(from, to.key())
+            .put_object_stream(&mut local_file, to.key())
            .await
            .with_context(|| format!("Failed to create s3 object with key {}", to.key()))?;
        if code != 200 {
--- a/pageserver/src/relish_storage/synced_storage.rs
+++ b/pageserver/src/relish_storage/synced_storage.rs
@@ -1,7 +1,6 @@
 use std::time::Duration;
 use std::{collections::BinaryHeap, sync::Mutex, thread};

-use crate::tenant_mgr;
 use crate::{relish_storage::RelishStorage, PageServerConf};

 lazy_static::lazy_static! {
@@ -32,26 +31,22 @@ pub fn run_storage_sync_thread<

    let handle = thread::Builder::new()
        .name("Queue based relish storage sync".to_string())
-        .spawn(move || {
-            while !tenant_mgr::shutdown_requested() {
-                let mut queue_accessor = UPLOAD_QUEUE.lock().unwrap();
-                log::debug!("Upload queue length: {}", queue_accessor.len());
-                let next_task = queue_accessor.pop();
-                drop(queue_accessor);
-                match next_task {
-                    Some(task) => runtime.block_on(async {
-                        // suppress warnings
-                        let _ = (config, task, &relish_storage, max_concurrent_sync);
-                        todo!("omitted for brevity")
-                    }),
-                    None => {
-                        thread::sleep(Duration::from_secs(1));
-                        continue;
-                    }
+        .spawn(move || loop {
+            let mut queue_accessor = UPLOAD_QUEUE.lock().unwrap();
+            log::debug!("Upload queue length: {}", queue_accessor.len());
+            let next_task = queue_accessor.pop();
+            drop(queue_accessor);
+            match next_task {
+                Some(task) => runtime.block_on(async {
+                    // suppress warnings
+                    let _ = (config, task, &relish_storage, max_concurrent_sync);
+                    todo!("omitted for brevity")
+                }),
+                None => {
+                    thread::sleep(Duration::from_secs(1));
+                    continue;
                }
            }
-            log::debug!("Queue based relish storage sync thread shut down");
-            Ok(())
        })?;
    Ok(Some(handle))
 }
--- a/pageserver/src/repository.rs
+++ b/pageserver/src/repository.rs
@@ -3,7 +3,7 @@ use anyhow::Result;
 use bytes::{Buf, BufMut, Bytes, BytesMut};
 use serde::{Deserialize, Serialize};
 use std::collections::HashSet;
-use std::ops::{AddAssign, Deref};
+use std::ops::AddAssign;
 use std::sync::Arc;
 use std::time::Duration;
 use zenith_utils::lsn::{Lsn, RecordLsn};
@@ -13,8 +13,6 @@ use zenith_utils::zid::ZTimelineId;
 /// A repository corresponds to one .zenith directory. One repository holds multiple
 /// timelines, forked off from the same initial call to 'initdb'.
 pub trait Repository: Send + Sync {
-    fn shutdown(&self) -> Result<()>;
-
    /// Get Timeline handle for given zenith timeline ID.
    fn get_timeline(&self, timelineid: ZTimelineId) -> Result<Arc<dyn Timeline>>;

@@ -125,6 +123,26 @@ pub trait Timeline: Send + Sync {
    // These are called by the WAL receiver to digest WAL records.
    //------------------------------------------------------------------------------

+    /// Put a new page version that can be constructed from a WAL record
+    ///
+    /// This will implicitly extend the relation, if the page is beyond the
+    /// current end-of-file.
+    fn put_wal_record(&self, tag: RelishTag, blknum: u32, rec: WALRecord) -> Result<()>;
+
+    /// Like put_wal_record, but with ready-made image of the page.
+    fn put_page_image(&self, tag: RelishTag, blknum: u32, lsn: Lsn, img: Bytes) -> Result<()>;
+
+    /// Truncate relation
+    fn put_truncation(&self, rel: RelishTag, lsn: Lsn, nblocks: u32) -> Result<()>;
+
+    /// This method is used for marking dropped relations and truncated SLRU files and aborted two phase records
+    fn drop_relish(&self, tag: RelishTag, lsn: Lsn) -> Result<()>;
+
+    /// Track end of the latest digested WAL record.
+    ///
+    /// Advance requires aligned LSN as an argument and would wake wait_lsn() callers.
+    /// Previous last record LSN is stored alongside the latest and can be read.
+    fn advance_last_record_lsn(&self, lsn: Lsn);
    /// Atomically get both last and prev.
    fn get_last_record_rlsn(&self) -> RecordLsn;
    /// Get last or prev record separately. Same as get_last_record_rlsn().last/prev.
@@ -132,9 +150,6 @@ pub trait Timeline: Send + Sync {
    fn get_prev_record_lsn(&self) -> Lsn;
    fn get_start_lsn(&self) -> Lsn;

-    /// Mutate the timeline with a [`TimelineWriter`].
-    fn writer<'a>(&'a self) -> Box<dyn TimelineWriter + 'a>;
-
    ///
    /// Flush to disk all data that was written with the put_* functions
    ///
@@ -153,35 +168,9 @@ pub trait Timeline: Send + Sync {
    fn get_current_logical_size_non_incremental(&self, lsn: Lsn) -> Result<usize>;
 }

-/// Various functions to mutate the timeline.
-// TODO Currently, Deref is used to allow easy access to read methods from this trait.
-// This is probably considered a bad practice in Rust and should be fixed eventually,
-// but will cause large code changes.
-pub trait TimelineWriter: Deref<Target = dyn Timeline> {
-    /// Put a new page version that can be constructed from a WAL record
-    ///
-    /// This will implicitly extend the relation, if the page is beyond the
-    /// current end-of-file.
-    fn put_wal_record(&self, lsn: Lsn, tag: RelishTag, blknum: u32, rec: WALRecord) -> Result<()>;
-
-    /// Like put_wal_record, but with ready-made image of the page.
-    fn put_page_image(&self, tag: RelishTag, blknum: u32, lsn: Lsn, img: Bytes) -> Result<()>;
-
-    /// Truncate relation
-    fn put_truncation(&self, rel: RelishTag, lsn: Lsn, nblocks: u32) -> Result<()>;
-
-    /// This method is used for marking dropped relations and truncated SLRU files and aborted two phase records
-    fn drop_relish(&self, tag: RelishTag, lsn: Lsn) -> Result<()>;
-
-    /// Track end of the latest digested WAL record.
-    ///
-    /// Advance requires aligned LSN as an argument and would wake wait_lsn() callers.
-    /// Previous last record LSN is stored alongside the latest and can be read.
-    fn advance_last_record_lsn(&self, lsn: Lsn);
-}
-
 #[derive(Debug, Clone, Serialize, Deserialize, PartialEq, Eq)]
 pub struct WALRecord {
+    pub lsn: Lsn, // LSN at the *end* of the record
    pub will_init: bool,
    pub rec: Bytes,
    // Remember the offset of main_data in rec,
@@ -192,19 +181,22 @@ pub struct WALRecord {

 impl WALRecord {
    pub fn pack(&self, buf: &mut BytesMut) {
+        buf.put_u64(self.lsn.0);
        buf.put_u8(self.will_init as u8);
        buf.put_u32(self.main_data_offset);
        buf.put_u32(self.rec.len() as u32);
        buf.put_slice(&self.rec[..]);
    }
    pub fn unpack(buf: &mut Bytes) -> WALRecord {
+        let lsn = Lsn::from(buf.get_u64());
        let will_init = buf.get_u8() != 0;
        let main_data_offset = buf.get_u32();
-        let rec_len = buf.get_u32() as usize;
-        let rec = buf.split_to(rec_len);
+        let mut dst = vec![0u8; buf.get_u32() as usize];
+        buf.copy_to_slice(&mut dst);
        WALRecord {
+            lsn,
            will_init,
-            rec,
+            rec: Bytes::from(dst),
            main_data_offset,
        }
    }
@@ -314,15 +306,14 @@ mod tests {

        // Create timeline to work on
        let tline = repo.create_empty_timeline(TIMELINE_ID)?;
-        let writer = tline.writer();

-        writer.put_page_image(TESTREL_A, 0, Lsn(0x20), TEST_IMG("foo blk 0 at 2"))?;
-        writer.put_page_image(TESTREL_A, 0, Lsn(0x20), TEST_IMG("foo blk 0 at 2"))?;
-        writer.put_page_image(TESTREL_A, 0, Lsn(0x30), TEST_IMG("foo blk 0 at 3"))?;
-        writer.put_page_image(TESTREL_A, 1, Lsn(0x40), TEST_IMG("foo blk 1 at 4"))?;
-        writer.put_page_image(TESTREL_A, 2, Lsn(0x50), TEST_IMG("foo blk 2 at 5"))?;
+        tline.put_page_image(TESTREL_A, 0, Lsn(0x20), TEST_IMG("foo blk 0 at 2"))?;
+        tline.put_page_image(TESTREL_A, 0, Lsn(0x20), TEST_IMG("foo blk 0 at 2"))?;
+        tline.put_page_image(TESTREL_A, 0, Lsn(0x30), TEST_IMG("foo blk 0 at 3"))?;
+        tline.put_page_image(TESTREL_A, 1, Lsn(0x40), TEST_IMG("foo blk 1 at 4"))?;
+        tline.put_page_image(TESTREL_A, 2, Lsn(0x50), TEST_IMG("foo blk 2 at 5"))?;

-        writer.advance_last_record_lsn(Lsn(0x50));
+        tline.advance_last_record_lsn(Lsn(0x50));

        assert_current_logical_size(&tline, Lsn(0x50));

@@ -368,8 +359,8 @@ mod tests {
        );

        // Truncate last block
-        writer.put_truncation(TESTREL_A, Lsn(0x60), 2)?;
-        writer.advance_last_record_lsn(Lsn(0x60));
+        tline.put_truncation(TESTREL_A, Lsn(0x60), 2)?;
+        tline.advance_last_record_lsn(Lsn(0x60));
        assert_current_logical_size(&tline, Lsn(0x60));

        // Check reported size and contents after truncation
@@ -391,13 +382,13 @@ mod tests {
        );

        // Truncate to zero length
-        writer.put_truncation(TESTREL_A, Lsn(0x68), 0)?;
-        writer.advance_last_record_lsn(Lsn(0x68));
+        tline.put_truncation(TESTREL_A, Lsn(0x68), 0)?;
+        tline.advance_last_record_lsn(Lsn(0x68));
        assert_eq!(tline.get_relish_size(TESTREL_A, Lsn(0x68))?.unwrap(), 0);

        // Extend from 0 to 2 blocks, leaving a gap
-        writer.put_page_image(TESTREL_A, 1, Lsn(0x70), TEST_IMG("foo blk 1"))?;
-        writer.advance_last_record_lsn(Lsn(0x70));
+        tline.put_page_image(TESTREL_A, 1, Lsn(0x70), TEST_IMG("foo blk 1"))?;
+        tline.advance_last_record_lsn(Lsn(0x70));
        assert_eq!(tline.get_relish_size(TESTREL_A, Lsn(0x70))?.unwrap(), 2);
        assert_eq!(tline.get_page_at_lsn(TESTREL_A, 0, Lsn(0x70))?, ZERO_PAGE);
        assert_eq!(
@@ -432,26 +423,25 @@ mod tests {

        // Create timeline to work on
        let tline = repo.create_empty_timeline(TIMELINE_ID)?;
-        let writer = tline.writer();

-        writer.put_page_image(TESTREL_A, 0, Lsn(0x20), TEST_IMG("foo blk 0 at 2"))?;
-        writer.advance_last_record_lsn(Lsn(0x20));
+        tline.put_page_image(TESTREL_A, 0, Lsn(0x20), TEST_IMG("foo blk 0 at 2"))?;
+        tline.advance_last_record_lsn(Lsn(0x20));

        // Check that rel exists and size is correct
        assert_eq!(tline.get_rel_exists(TESTREL_A, Lsn(0x20))?, true);
        assert_eq!(tline.get_relish_size(TESTREL_A, Lsn(0x20))?.unwrap(), 1);

        // Drop relish
-        writer.drop_relish(TESTREL_A, Lsn(0x30))?;
-        writer.advance_last_record_lsn(Lsn(0x30));
+        tline.drop_relish(TESTREL_A, Lsn(0x30))?;
+        tline.advance_last_record_lsn(Lsn(0x30));

        // Check that rel is not visible anymore
        assert_eq!(tline.get_rel_exists(TESTREL_A, Lsn(0x30))?, false);
        assert!(tline.get_relish_size(TESTREL_A, Lsn(0x30))?.is_none());

        // Extend it again
-        writer.put_page_image(TESTREL_A, 0, Lsn(0x40), TEST_IMG("foo blk 0 at 4"))?;
-        writer.advance_last_record_lsn(Lsn(0x40));
+        tline.put_page_image(TESTREL_A, 0, Lsn(0x40), TEST_IMG("foo blk 0 at 4"))?;
+        tline.advance_last_record_lsn(Lsn(0x40));

        // Check that rel exists and size is correct
        assert_eq!(tline.get_rel_exists(TESTREL_A, Lsn(0x40))?, true);
@@ -469,7 +459,6 @@ mod tests {

        // Create timeline to work on
        let tline = repo.create_empty_timeline(TIMELINE_ID)?;
-        let writer = tline.writer();

        //from storage_layer.rs
        const RELISH_SEG_SIZE: u32 = 10 * 1024 * 1024 / 8192;
@@ -479,10 +468,10 @@ mod tests {
        for blkno in 0..relsize {
            let lsn = Lsn(0x20);
            let data = format!("foo blk {} at {}", blkno, lsn);
-            writer.put_page_image(TESTREL_A, blkno, lsn, TEST_IMG(&data))?;
+            tline.put_page_image(TESTREL_A, blkno, lsn, TEST_IMG(&data))?;
        }

-        writer.advance_last_record_lsn(Lsn(0x20));
+        tline.advance_last_record_lsn(Lsn(0x20));

        // The relation was created at LSN 2, not visible at LSN 1 yet.
        assert_eq!(tline.get_rel_exists(TESTREL_A, Lsn(0x10))?, false);
@@ -506,8 +495,8 @@ mod tests {

        // Truncate relation so that second segment was dropped
        // - only leave one page
-        writer.put_truncation(TESTREL_A, Lsn(0x60), 1)?;
-        writer.advance_last_record_lsn(Lsn(0x60));
+        tline.put_truncation(TESTREL_A, Lsn(0x60), 1)?;
+        tline.advance_last_record_lsn(Lsn(0x60));

        // Check reported size and contents after truncation
        assert_eq!(tline.get_relish_size(TESTREL_A, Lsn(0x60))?.unwrap(), 1);
@@ -540,9 +529,9 @@ mod tests {
        for blkno in 0..relsize {
            let lsn = Lsn(0x80);
            let data = format!("foo blk {} at {}", blkno, lsn);
-            writer.put_page_image(TESTREL_A, blkno, lsn, TEST_IMG(&data))?;
+            tline.put_page_image(TESTREL_A, blkno, lsn, TEST_IMG(&data))?;
        }
-        writer.advance_last_record_lsn(Lsn(0x80));
+        tline.advance_last_record_lsn(Lsn(0x80));

        assert_eq!(tline.get_rel_exists(TESTREL_A, Lsn(0x80))?, true);
        assert_eq!(
@@ -568,15 +557,14 @@ mod tests {
    fn test_large_rel() -> Result<()> {
        let repo = RepoHarness::create("test_large_rel")?.load();
        let tline = repo.create_empty_timeline(TIMELINE_ID)?;
-        let writer = tline.writer();

        let mut lsn = 0x10;
        for blknum in 0..pg_constants::RELSEG_SIZE + 1 {
            let img = TEST_IMG(&format!("foo blk {} at {}", blknum, Lsn(lsn)));
            lsn += 0x10;
-            writer.put_page_image(TESTREL_A, blknum as u32, Lsn(lsn), img)?;
+            tline.put_page_image(TESTREL_A, blknum as u32, Lsn(lsn), img)?;
        }
-        writer.advance_last_record_lsn(Lsn(lsn));
+        tline.advance_last_record_lsn(Lsn(lsn));

        assert_current_logical_size(&tline, Lsn(lsn));

@@ -587,8 +575,8 @@ mod tests {

        // Truncate one block
        lsn += 0x10;
-        writer.put_truncation(TESTREL_A, Lsn(lsn), pg_constants::RELSEG_SIZE)?;
-        writer.advance_last_record_lsn(Lsn(lsn));
+        tline.put_truncation(TESTREL_A, Lsn(lsn), pg_constants::RELSEG_SIZE)?;
+        tline.advance_last_record_lsn(Lsn(lsn));
        assert_eq!(
            tline.get_relish_size(TESTREL_A, Lsn(lsn))?.unwrap(),
            pg_constants::RELSEG_SIZE
@@ -597,8 +585,8 @@ mod tests {

        // Truncate another block
        lsn += 0x10;
-        writer.put_truncation(TESTREL_A, Lsn(lsn), pg_constants::RELSEG_SIZE - 1)?;
-        writer.advance_last_record_lsn(Lsn(lsn));
+        tline.put_truncation(TESTREL_A, Lsn(lsn), pg_constants::RELSEG_SIZE - 1)?;
+        tline.advance_last_record_lsn(Lsn(lsn));
        assert_eq!(
            tline.get_relish_size(TESTREL_A, Lsn(lsn))?.unwrap(),
            pg_constants::RELSEG_SIZE - 1
@@ -610,8 +598,8 @@ mod tests {
        let mut size: i32 = 3000;
        while size >= 0 {
            lsn += 0x10;
-            writer.put_truncation(TESTREL_A, Lsn(lsn), size as u32)?;
-            writer.advance_last_record_lsn(Lsn(lsn));
+            tline.put_truncation(TESTREL_A, Lsn(lsn), size as u32)?;
+            tline.advance_last_record_lsn(Lsn(lsn));
            assert_eq!(
                tline.get_relish_size(TESTREL_A, Lsn(lsn))?.unwrap(),
                size as u32
@@ -631,17 +619,16 @@ mod tests {
    fn test_list_rels_drop() -> Result<()> {
        let repo = RepoHarness::create("test_list_rels_drop")?.load();
        let tline = repo.create_empty_timeline(TIMELINE_ID)?;
-        let writer = tline.writer();
        const TESTDB: u32 = 111;

        // Import initial dummy checkpoint record, otherwise the get_timeline() call
        // after branching fails below
-        writer.put_page_image(RelishTag::Checkpoint, 0, Lsn(0x10), ZERO_CHECKPOINT.clone())?;
+        tline.put_page_image(RelishTag::Checkpoint, 0, Lsn(0x10), ZERO_CHECKPOINT.clone())?;

        // Create a relation on the timeline
-        writer.put_page_image(TESTREL_A, 0, Lsn(0x20), TEST_IMG("foo blk 0 at 2"))?;
+        tline.put_page_image(TESTREL_A, 0, Lsn(0x20), TEST_IMG("foo blk 0 at 2"))?;

-        writer.advance_last_record_lsn(Lsn(0x30));
+        tline.advance_last_record_lsn(Lsn(0x30));

        // Check that list_rels() lists it after LSN 2, but no before it
        assert!(!tline.list_rels(0, TESTDB, Lsn(0x10))?.contains(&TESTREL_A));
@@ -651,17 +638,14 @@ mod tests {
        // Create a branch, check that the relation is visible there
        repo.branch_timeline(TIMELINE_ID, NEW_TIMELINE_ID, Lsn(0x30))?;
        let newtline = repo.get_timeline(NEW_TIMELINE_ID)?;
-        let new_writer = newtline.writer();

        assert!(newtline
            .list_rels(0, TESTDB, Lsn(0x30))?
            .contains(&TESTREL_A));

        // Drop it on the branch
-        new_writer.drop_relish(TESTREL_A, Lsn(0x40))?;
-        new_writer.advance_last_record_lsn(Lsn(0x40));
-
-        drop(new_writer);
+        newtline.drop_relish(TESTREL_A, Lsn(0x40))?;
+        newtline.advance_last_record_lsn(Lsn(0x40));

        // Check that it's no longer listed on the branch after the point where it was dropped
        assert!(newtline
@@ -689,30 +673,28 @@ mod tests {
    fn test_branch() -> Result<()> {
        let repo = RepoHarness::create("test_branch")?.load();
        let tline = repo.create_empty_timeline(TIMELINE_ID)?;
-        let writer = tline.writer();

        // Import initial dummy checkpoint record, otherwise the get_timeline() call
        // after branching fails below
-        writer.put_page_image(RelishTag::Checkpoint, 0, Lsn(0x10), ZERO_CHECKPOINT.clone())?;
+        tline.put_page_image(RelishTag::Checkpoint, 0, Lsn(0x10), ZERO_CHECKPOINT.clone())?;

        // Create a relation on the timeline
-        writer.put_page_image(TESTREL_A, 0, Lsn(0x20), TEST_IMG("foo blk 0 at 2"))?;
-        writer.put_page_image(TESTREL_A, 0, Lsn(0x30), TEST_IMG("foo blk 0 at 3"))?;
-        writer.put_page_image(TESTREL_A, 0, Lsn(0x40), TEST_IMG("foo blk 0 at 4"))?;
+        tline.put_page_image(TESTREL_A, 0, Lsn(0x20), TEST_IMG("foo blk 0 at 2"))?;
+        tline.put_page_image(TESTREL_A, 0, Lsn(0x30), TEST_IMG("foo blk 0 at 3"))?;
+        tline.put_page_image(TESTREL_A, 0, Lsn(0x40), TEST_IMG("foo blk 0 at 4"))?;

        // Create another relation
-        writer.put_page_image(TESTREL_B, 0, Lsn(0x20), TEST_IMG("foobar blk 0 at 2"))?;
+        tline.put_page_image(TESTREL_B, 0, Lsn(0x20), TEST_IMG("foobar blk 0 at 2"))?;

-        writer.advance_last_record_lsn(Lsn(0x40));
+        tline.advance_last_record_lsn(Lsn(0x40));
        assert_current_logical_size(&tline, Lsn(0x40));

        // Branch the history, modify relation differently on the new timeline
        repo.branch_timeline(TIMELINE_ID, NEW_TIMELINE_ID, Lsn(0x30))?;
        let newtline = repo.get_timeline(NEW_TIMELINE_ID)?;
-        let new_writer = newtline.writer();

-        new_writer.put_page_image(TESTREL_A, 0, Lsn(0x40), TEST_IMG("bar blk 0 at 4"))?;
-        new_writer.advance_last_record_lsn(Lsn(0x40));
+        newtline.put_page_image(TESTREL_A, 0, Lsn(0x40), TEST_IMG("bar blk 0 at 4"))?;
+        newtline.advance_last_record_lsn(Lsn(0x40));

        // Check page contents on both branches
        assert_eq!(
@@ -828,7 +810,7 @@ mod tests {
            blknum: u32,
            lsn: Lsn,
            base_img: Option<Bytes>,
-            records: Vec<(Lsn, WALRecord)>,
+            records: Vec<WALRecord>,
        ) -> Result<Bytes, WalRedoError> {
            let s = format!(
                "redo for {} blk {} to get to {}, with {} and {} records",
--- a/pageserver/src/restore_local_repo.rs
+++ b/pageserver/src/restore_local_repo.rs
@@ -2,6 +2,7 @@
 //! Import data and WAL from a PostgreSQL data directory and WAL segments into
 //! zenith Timeline.
 //!
+use log::*;
 use postgres_ffi::nonrelfile_utils::clogpage_precedes;
 use postgres_ffi::nonrelfile_utils::slru_may_delete_clogsegment;
 use std::cmp::min;
@@ -12,7 +13,6 @@ use std::path::Path;

 use anyhow::{bail, Result};
 use bytes::{Buf, Bytes};
-use tracing::*;

 use crate::relish::*;
 use crate::repository::*;
@@ -34,7 +34,7 @@ static ZERO_PAGE: Bytes = Bytes::from_static(&[0u8; 8192]);
 ///
 pub fn import_timeline_from_postgres_datadir(
    path: &Path,
-    writer: &dyn TimelineWriter,
+    timeline: &dyn Timeline,
    lsn: Lsn,
 ) -> Result<()> {
    // Scan 'global'
@@ -44,10 +44,10 @@ pub fn import_timeline_from_postgres_datadir(
            None => continue,

            Some("pg_control") => {
-                import_control_file(writer, lsn, &direntry.path())?;
+                import_control_file(timeline, lsn, &direntry.path())?;
            }
            Some("pg_filenode.map") => import_nonrel_file(
-                writer,
+                timeline,
                lsn,
                RelishTag::FileNodeMap {
                    spcnode: pg_constants::GLOBALTABLESPACE_OID,
@@ -59,7 +59,7 @@ pub fn import_timeline_from_postgres_datadir(
            // Load any relation files into the page server
            _ => import_relfile(
                &direntry.path(),
-                writer,
+                timeline,
                lsn,
                pg_constants::GLOBALTABLESPACE_OID,
                0,
@@ -86,7 +86,7 @@ pub fn import_timeline_from_postgres_datadir(

                Some("PG_VERSION") => continue,
                Some("pg_filenode.map") => import_nonrel_file(
-                    writer,
+                    timeline,
                    lsn,
                    RelishTag::FileNodeMap {
                        spcnode: pg_constants::DEFAULTTABLESPACE_OID,
@@ -98,7 +98,7 @@ pub fn import_timeline_from_postgres_datadir(
                // Load any relation files into the page server
                _ => import_relfile(
                    &direntry.path(),
-                    writer,
+                    timeline,
                    lsn,
                    pg_constants::DEFAULTTABLESPACE_OID,
                    dboid,
@@ -108,24 +108,24 @@ pub fn import_timeline_from_postgres_datadir(
    }
    for entry in fs::read_dir(path.join("pg_xact"))? {
        let entry = entry?;
-        import_slru_file(writer, lsn, SlruKind::Clog, &entry.path())?;
+        import_slru_file(timeline, lsn, SlruKind::Clog, &entry.path())?;
    }
    for entry in fs::read_dir(path.join("pg_multixact").join("members"))? {
        let entry = entry?;
-        import_slru_file(writer, lsn, SlruKind::MultiXactMembers, &entry.path())?;
+        import_slru_file(timeline, lsn, SlruKind::MultiXactMembers, &entry.path())?;
    }
    for entry in fs::read_dir(path.join("pg_multixact").join("offsets"))? {
        let entry = entry?;
-        import_slru_file(writer, lsn, SlruKind::MultiXactOffsets, &entry.path())?;
+        import_slru_file(timeline, lsn, SlruKind::MultiXactOffsets, &entry.path())?;
    }
    for entry in fs::read_dir(path.join("pg_twophase"))? {
        let entry = entry?;
        let xid = u32::from_str_radix(entry.path().to_str().unwrap(), 16)?;
-        import_nonrel_file(writer, lsn, RelishTag::TwoPhase { xid }, &entry.path())?;
+        import_nonrel_file(timeline, lsn, RelishTag::TwoPhase { xid }, &entry.path())?;
    }
    // TODO: Scan pg_tblspc

-    writer.advance_last_record_lsn(lsn);
+    timeline.advance_last_record_lsn(lsn);

    Ok(())
 }
@@ -133,13 +133,12 @@ pub fn import_timeline_from_postgres_datadir(
 // subroutine of import_timeline_from_postgres_datadir(), to load one relation file.
 fn import_relfile(
    path: &Path,
-    timeline: &dyn TimelineWriter,
+    timeline: &dyn Timeline,
    lsn: Lsn,
    spcoid: Oid,
    dboid: Oid,
 ) -> Result<()> {
    // Does it look like a relation file?
-    trace!("importing rel file {}", path.display());

    let p = parse_relfilename(path.file_name().unwrap().to_str().unwrap());
    if let Err(e) = p {
@@ -167,14 +166,14 @@ fn import_relfile(
            }

            // TODO: UnexpectedEof is expected
-            Err(err) => match err.kind() {
+            Err(e) => match e.kind() {
                std::io::ErrorKind::UnexpectedEof => {
                    // reached EOF. That's expected.
                    // FIXME: maybe check that we read the full length of the file?
                    break;
                }
                _ => {
-                    bail!("error reading file {}: {:#}", path.display(), err);
+                    bail!("error reading file {}: {:#}", path.display(), e);
                }
            },
        };
@@ -191,7 +190,7 @@ fn import_relfile(
 /// are just slurped into the repository as one blob.
 ///
 fn import_nonrel_file(
-    timeline: &dyn TimelineWriter,
+    timeline: &dyn Timeline,
    lsn: Lsn,
    tag: RelishTag,
    path: &Path,
@@ -201,7 +200,7 @@ fn import_nonrel_file(
    // read the whole file
    file.read_to_end(&mut buffer)?;

-    trace!("importing non-rel file {}", path.display());
+    info!("importing non-rel file {}", path.display());

    timeline.put_page_image(tag, 0, lsn, Bytes::copy_from_slice(&buffer[..]))?;
    Ok(())
@@ -212,13 +211,13 @@ fn import_nonrel_file(
 ///
 /// The control file is imported as is, but we also extract the checkpoint record
 /// from it and store it separated.
-fn import_control_file(timeline: &dyn TimelineWriter, lsn: Lsn, path: &Path) -> Result<()> {
+fn import_control_file(timeline: &dyn Timeline, lsn: Lsn, path: &Path) -> Result<()> {
    let mut file = File::open(path)?;
    let mut buffer = Vec::new();
    // read the whole file
    file.read_to_end(&mut buffer)?;

-    trace!("importing control file {}", path.display());
+    info!("importing control file {}", path.display());

    // Import it as ControlFile
    timeline.put_page_image(
@@ -239,18 +238,13 @@ fn import_control_file(timeline: &dyn TimelineWriter, lsn: Lsn, path: &Path) ->
 ///
 /// Import an SLRU segment file
 ///
-fn import_slru_file(
-    timeline: &dyn TimelineWriter,
-    lsn: Lsn,
-    slru: SlruKind,
-    path: &Path,
-) -> Result<()> {
+fn import_slru_file(timeline: &dyn Timeline, lsn: Lsn, slru: SlruKind, path: &Path) -> Result<()> {
    // Does it look like an SLRU file?
    let mut file = File::open(path)?;
    let mut buf: [u8; 8192] = [0u8; 8192];
    let segno = u32::from_str_radix(path.file_name().unwrap().to_str().unwrap(), 16)?;

-    trace!("importing slru file {}", path.display());
+    info!("importing slru file {}", path.display());

    let mut rpageno = 0;
    loop {
@@ -266,14 +260,14 @@ fn import_slru_file(
            }

            // TODO: UnexpectedEof is expected
-            Err(err) => match err.kind() {
+            Err(e) => match e.kind() {
                std::io::ErrorKind::UnexpectedEof => {
                    // reached EOF. That's expected.
                    // FIXME: maybe check that we read the full length of the file?
                    break;
                }
                _ => {
-                    bail!("error reading file {}: {:#}", path.display(), err);
+                    bail!("error reading file {}: {:#}", path.display(), e);
                }
            },
        };
@@ -291,15 +285,12 @@ fn import_slru_file(
 ///
 pub fn save_decoded_record(
    checkpoint: &mut CheckPoint,
-    checkpoint_modified: &mut bool,
-    timeline: &dyn TimelineWriter,
+    timeline: &dyn Timeline,
    decoded: &DecodedWALRecord,
    recdata: Bytes,
    lsn: Lsn,
 ) -> Result<()> {
-    if checkpoint.update_next_xid(decoded.xl_xid) {
-        *checkpoint_modified = true;
-    }
+    checkpoint.update_next_xid(decoded.xl_xid);

    // Iterate through all the blocks that the record modifies, and
    // "put" a separate copy of the record for each block.
@@ -312,12 +303,13 @@ pub fn save_decoded_record(
        });

        let rec = WALRecord {
+            lsn,
            will_init: blk.will_init || blk.apply_image,
            rec: recdata.clone(),
            main_data_offset: decoded.main_data_offset as u32,
        };

-        timeline.put_wal_record(lsn, tag, blk.blkno, rec)?;
+        timeline.put_wal_record(tag, blk.blkno, rec)?;
    }

    let mut buf = decoded.record.clone();
@@ -382,7 +374,7 @@ pub fn save_decoded_record(
        } else {
            assert!(info == pg_constants::CLOG_TRUNCATE);
            let xlrec = XlClogTruncate::decode(&mut buf);
-            save_clog_truncate_record(checkpoint, checkpoint_modified, timeline, lsn, &xlrec)?;
+            save_clog_truncate_record(checkpoint, timeline, lsn, &xlrec)?;
        }
    } else if decoded.xl_rmid == pg_constants::RM_XACT_ID {
        let info = decoded.xl_info & pg_constants::XLOG_XACT_OPMASK;
@@ -451,17 +443,10 @@ pub fn save_decoded_record(
            )?;
        } else if info == pg_constants::XLOG_MULTIXACT_CREATE_ID {
            let xlrec = XlMultiXactCreate::decode(&mut buf);
-            save_multixact_create_record(
-                checkpoint,
-                checkpoint_modified,
-                timeline,
-                lsn,
-                &xlrec,
-                decoded,
-            )?;
+            save_multixact_create_record(checkpoint, timeline, lsn, &xlrec, decoded)?;
        } else if info == pg_constants::XLOG_MULTIXACT_TRUNCATE_ID {
            let xlrec = XlMultiXactTruncate::decode(&mut buf);
-            save_multixact_truncate_record(checkpoint, checkpoint_modified, timeline, lsn, &xlrec)?;
+            save_multixact_truncate_record(checkpoint, timeline, lsn, &xlrec)?;
        }
    } else if decoded.xl_rmid == pg_constants::RM_RELMAP_ID {
        let xlrec = XlRelmapUpdate::decode(&mut buf);
@@ -470,10 +455,7 @@ pub fn save_decoded_record(
        let info = decoded.xl_info & pg_constants::XLR_RMGR_INFO_MASK;
        if info == pg_constants::XLOG_NEXTOID {
            let next_oid = buf.get_u32_le();
-            if checkpoint.nextOid != next_oid {
-                checkpoint.nextOid = next_oid;
-                *checkpoint_modified = true;
-            }
+            checkpoint.nextOid = next_oid;
        } else if info == pg_constants::XLOG_CHECKPOINT_ONLINE
            || info == pg_constants::XLOG_CHECKPOINT_SHUTDOWN
        {
@@ -489,7 +471,6 @@ pub fn save_decoded_record(
            );
            if (checkpoint.oldestXid.wrapping_sub(xlog_checkpoint.oldestXid) as i32) < 0 {
                checkpoint.oldestXid = xlog_checkpoint.oldestXid;
-                *checkpoint_modified = true;
            }
        }
    }
@@ -497,11 +478,7 @@ pub fn save_decoded_record(
 }

 /// Subroutine of save_decoded_record(), to handle an XLOG_DBASE_CREATE record.
-fn save_xlog_dbase_create(
-    timeline: &dyn TimelineWriter,
-    lsn: Lsn,
-    rec: &XlCreateDatabase,
-) -> Result<()> {
+fn save_xlog_dbase_create(timeline: &dyn Timeline, lsn: Lsn, rec: &XlCreateDatabase) -> Result<()> {
    let db_id = rec.db_id;
    let tablespace_id = rec.tablespace_id;
    let src_db_id = rec.src_db_id;
@@ -578,11 +555,7 @@ fn save_xlog_dbase_create(
 /// Subroutine of save_decoded_record(), to handle an XLOG_SMGR_TRUNCATE record.
 ///
 /// This is the same logic as in PostgreSQL's smgr_redo() function.
-fn save_xlog_smgr_truncate(
-    timeline: &dyn TimelineWriter,
-    lsn: Lsn,
-    rec: &XlSmgrTruncate,
-) -> Result<()> {
+fn save_xlog_smgr_truncate(timeline: &dyn Timeline, lsn: Lsn, rec: &XlSmgrTruncate) -> Result<()> {
    let spcnode = rec.rnode.spcnode;
    let dbnode = rec.rnode.dbnode;
    let relnode = rec.rnode.relnode;
@@ -644,7 +617,7 @@ fn save_xlog_smgr_truncate(
 /// Subroutine of save_decoded_record(), to handle an XLOG_XACT_* records.
 ///
 fn save_xact_record(
-    timeline: &dyn TimelineWriter,
+    timeline: &dyn Timeline,
    lsn: Lsn,
    parsed: &XlXactParsedRecord,
    decoded: &DecodedWALRecord,
@@ -655,12 +628,12 @@ fn save_xact_record(
    let segno = pageno / pg_constants::SLRU_PAGES_PER_SEGMENT;
    let rpageno = pageno % pg_constants::SLRU_PAGES_PER_SEGMENT;
    let rec = WALRecord {
+        lsn,
        will_init: false,
        rec: decoded.record.clone(),
        main_data_offset: decoded.main_data_offset as u32,
    };
    timeline.put_wal_record(
-        lsn,
        RelishTag::Slru {
            slru: SlruKind::Clog,
            segno,
@@ -676,7 +649,6 @@ fn save_xact_record(
            let segno = pageno / pg_constants::SLRU_PAGES_PER_SEGMENT;
            let rpageno = pageno % pg_constants::SLRU_PAGES_PER_SEGMENT;
            timeline.put_wal_record(
-                lsn,
                RelishTag::Slru {
                    slru: SlruKind::Clog,
                    segno,
@@ -702,8 +674,7 @@ fn save_xact_record(

 fn save_clog_truncate_record(
    checkpoint: &mut CheckPoint,
-    checkpoint_modified: &mut bool,
-    timeline: &dyn TimelineWriter,
+    timeline: &dyn Timeline,
    lsn: Lsn,
    xlrec: &XlClogTruncate,
 ) -> Result<()> {
@@ -721,7 +692,6 @@ fn save_clog_truncate_record(
    // TODO Figure out if there will be any issues with replica.
    checkpoint.oldestXid = xlrec.oldest_xid;
    checkpoint.oldestXidDB = xlrec.oldest_xid_db;
-    *checkpoint_modified = true;

    // TODO Treat AdvanceOldestClogXid() or write a comment why we don't need it

@@ -764,13 +734,13 @@ fn save_clog_truncate_record(

 fn save_multixact_create_record(
    checkpoint: &mut CheckPoint,
-    checkpoint_modified: &mut bool,
-    timeline: &dyn TimelineWriter,
+    timeline: &dyn Timeline,
    lsn: Lsn,
    xlrec: &XlMultiXactCreate,
    decoded: &DecodedWALRecord,
 ) -> Result<()> {
    let rec = WALRecord {
+        lsn,
        will_init: false,
        rec: decoded.record.clone(),
        main_data_offset: decoded.main_data_offset as u32,
@@ -779,7 +749,6 @@ fn save_multixact_create_record(
    let segno = pageno / pg_constants::SLRU_PAGES_PER_SEGMENT;
    let rpageno = pageno % pg_constants::SLRU_PAGES_PER_SEGMENT;
    timeline.put_wal_record(
-        lsn,
        RelishTag::Slru {
            slru: SlruKind::MultiXactOffsets,
            segno,
@@ -799,7 +768,6 @@ fn save_multixact_create_record(
        let segno = pageno / pg_constants::SLRU_PAGES_PER_SEGMENT;
        let rpageno = pageno % pg_constants::SLRU_PAGES_PER_SEGMENT;
        timeline.put_wal_record(
-            lsn,
            RelishTag::Slru {
                slru: SlruKind::MultiXactMembers,
                segno,
@@ -822,11 +790,9 @@ fn save_multixact_create_record(
    }
    if xlrec.mid >= checkpoint.nextMulti {
        checkpoint.nextMulti = xlrec.mid + 1;
-        *checkpoint_modified = true;
    }
    if xlrec.moff + xlrec.nmembers > checkpoint.nextMultiOffset {
        checkpoint.nextMultiOffset = xlrec.moff + xlrec.nmembers;
-        *checkpoint_modified = true;
    }
    let max_mbr_xid = xlrec.members.iter().fold(0u32, |acc, mbr| {
        if mbr.xid.wrapping_sub(acc) as i32 > 0 {
@@ -836,22 +802,18 @@ fn save_multixact_create_record(
        }
    });

-    if checkpoint.update_next_xid(max_mbr_xid) {
-        *checkpoint_modified = true;
-    }
+    checkpoint.update_next_xid(max_mbr_xid);
    Ok(())
 }

 fn save_multixact_truncate_record(
    checkpoint: &mut CheckPoint,
-    checkpoint_modified: &mut bool,
-    timeline: &dyn TimelineWriter,
+    timeline: &dyn Timeline,
    lsn: Lsn,
    xlrec: &XlMultiXactTruncate,
 ) -> Result<()> {
    checkpoint.oldestMulti = xlrec.end_trunc_off;
    checkpoint.oldestMultiDB = xlrec.oldest_multi_db;
-    *checkpoint_modified = true;

    // PerformMembersTruncation
    let maxsegment: i32 = mx_offset_to_member_segment(pg_constants::MAX_MULTIXACT_OFFSET);
@@ -885,7 +847,7 @@ fn save_multixact_truncate_record(
 }

 fn save_relmap_page(
-    timeline: &dyn TimelineWriter,
+    timeline: &dyn Timeline,
    lsn: Lsn,
    xlrec: &XlRelmapUpdate,
    decoded: &DecodedWALRecord,
--- a/pageserver/src/tenant_mgr.rs
+++ b/pageserver/src/tenant_mgr.rs
@@ -8,14 +8,12 @@ use crate::walredo::PostgresRedoManager;
 use crate::PageServerConf;
 use anyhow::{anyhow, bail, Context, Result};
 use lazy_static::lazy_static;
-use log::{debug, info};
+use log::info;
 use std::collections::hash_map::Entry;
 use std::collections::HashMap;
 use std::fs;
 use std::str::FromStr;
-use std::sync::atomic::{AtomicBool, Ordering};
 use std::sync::{Arc, Mutex, MutexGuard};
-use std::thread::JoinHandle;
 use zenith_utils::zid::{ZTenantId, ZTimelineId};

 lazy_static! {
@@ -26,19 +24,6 @@ lazy_static! {
 fn access_repository() -> MutexGuard<'static, HashMap<ZTenantId, Arc<dyn Repository>>> {
    REPOSITORY.lock().unwrap()
 }
-struct TenantHandleEntry {
-    checkpointer_handle: Option<JoinHandle<()>>,
-    gc_handle: Option<JoinHandle<()>>,
-}
-
-// Logically these handles belong to Repository,
-// but it's just simpler to store them separately
-lazy_static! {
-    static ref TENANT_HANDLES: Mutex<HashMap<ZTenantId, TenantHandleEntry>> =
-        Mutex::new(HashMap::new());
-}
-
-static SHUTDOWN_REQUESTED: AtomicBool = AtomicBool::new(false);

 pub fn init(conf: &'static PageServerConf) {
    let mut m = access_repository();
@@ -62,18 +47,8 @@ fn init_repo(conf: &'static PageServerConf, tenant_id: ZTenantId) -> Arc<Layered
        tenant_id,
        true,
    ));
-
-    let checkpointer_handle = LayeredRepository::launch_checkpointer_thread(conf, repo.clone());
-    let gc_handle = LayeredRepository::launch_gc_thread(conf, repo.clone());
-
-    let mut handles = TENANT_HANDLES.lock().unwrap();
-    let h = TenantHandleEntry {
-        checkpointer_handle: Some(checkpointer_handle),
-        gc_handle: Some(gc_handle),
-    };
-
-    handles.insert(tenant_id, h);
-
+    LayeredRepository::launch_checkpointer_thread(conf, repo.clone());
+    LayeredRepository::launch_gc_thread(conf, repo.clone());
    repo
 }

@@ -107,35 +82,6 @@ fn init_timeline(repo: &dyn Repository, timeline_id: ZTimelineId) {
    }
 }

-// Check this flag in the thread loops to know when to exit
-pub fn shutdown_requested() -> bool {
-    SHUTDOWN_REQUESTED.load(Ordering::Relaxed)
-}
-
-pub fn stop_tenant_threads(tenantid: ZTenantId) {
-    let mut handles = TENANT_HANDLES.lock().unwrap();
-    if let Some(h) = handles.get_mut(&tenantid) {
-        h.checkpointer_handle.take().map(JoinHandle::join);
-        debug!("checkpointer for tenant {} has stopped", tenantid);
-        h.gc_handle.take().map(JoinHandle::join);
-        debug!("gc for tenant {} has stopped", tenantid);
-    }
-}
-
-pub fn shutdown_all_tenants() -> Result<()> {
-    SHUTDOWN_REQUESTED.swap(true, Ordering::Relaxed);
-
-    let tenants = list_tenants()?;
-    for tenantid in tenants {
-        stop_tenant_threads(tenantid);
-        let repo = get_repository_for_tenant(tenantid)?;
-        debug!("shutdown tenant {}", tenantid);
-        repo.shutdown()?;
-    }
-
-    Ok(())
-}
-
 pub fn create_repository_for_tenant(
    conf: &'static PageServerConf,
    tenantid: ZTenantId,
@@ -169,14 +115,3 @@ pub fn get_timeline_for_tenant(
        .get_timeline(timelineid)
        .with_context(|| format!("cannot fetch timeline {}", timelineid))
 }
-
-fn list_tenants() -> Result<Vec<ZTenantId>> {
-    let o = &mut REPOSITORY.lock().unwrap();
-
-    o.iter()
-        .map(|tenant| {
-            let (tenantid, _) = tenant;
-            Ok(*tenantid)
-        })
-        .collect()
-}
--- a/pageserver/src/waldecoder.rs
+++ b/pageserver/src/waldecoder.rs
@@ -72,10 +72,6 @@ impl WalStreamDecoder {
    ///     Err(WalDecodeError): an error occured while decoding, meaning the input was invalid.
    ///
    pub fn poll_decode(&mut self) -> Result<Option<(Lsn, Bytes)>, WalDecodeError> {
-        let recordbuf;
-
-        // Run state machine that validates page headers, and reassembles records
-        // that cross page boundaries.
        loop {
            // parse and verify page boundaries as we go
            if self.lsn.segment_offset(pg_constants::WAL_SEGMENT_SIZE) == 0 {
@@ -124,41 +120,29 @@ impl WalStreamDecoder {
                self.lsn += self.padlen as u64;
                self.padlen = 0;
            } else if self.contlen == 0 {
-                assert!(self.recordbuf.is_empty());
-
                // need to have at least the xl_tot_len field
+
                if self.inputbuf.remaining() < 4 {
                    return Ok(None);
                }

-                // peek xl_tot_len at the beginning of the record.
-                // FIXME: assumes little-endian
+                // read xl_tot_len FIXME: assumes little-endian
                self.startlsn = self.lsn;
-                let xl_tot_len = (&self.inputbuf[0..4]).get_u32_le();
+                let xl_tot_len = self.inputbuf.get_u32_le();
                if (xl_tot_len as usize) < XLOG_SIZE_OF_XLOG_RECORD {
                    return Err(WalDecodeError {
                        msg: format!("invalid xl_tot_len {}", xl_tot_len),
                        lsn: self.lsn,
                    });
                }
+                self.lsn += 4;

-                // Fast path for the common case that the whole record fits on the page.
-                let pageleft = self.lsn.remaining_in_block() as u32;
-                if self.inputbuf.remaining() >= xl_tot_len as usize && xl_tot_len <= pageleft {
-                    // Take the record from the 'inputbuf', and validate it.
-                    recordbuf = self.inputbuf.copy_to_bytes(xl_tot_len as usize);
-                    self.lsn += xl_tot_len as u64;
-                    break;
-                } else {
-                    // Need to assemble the record from pieces. Remember the size of the
-                    // record, and loop back. On next iteration, we will reach the 'else'
-                    // branch below, and copy the part of the record that was on this page
-                    // to 'recordbuf'.  Subsequent iterations will skip page headers, and
-                    // append the continuations from the next pages to 'recordbuf'.
-                    self.recordbuf.reserve(xl_tot_len as usize);
-                    self.contlen = xl_tot_len;
-                    continue;
-                }
+                self.recordbuf.clear();
+                self.recordbuf.reserve(xl_tot_len as usize);
+                self.recordbuf.put_u32_le(xl_tot_len);
+
+                self.contlen = xl_tot_len - 4;
+                continue;
            } else {
                // we're continuing a record, possibly from previous page.
                let pageleft = self.lsn.remaining_in_block() as u32;
@@ -175,42 +159,47 @@ impl WalStreamDecoder {
                self.contlen -= n as u32;

                if self.contlen == 0 {
-                    // The record is now complete.
-                    recordbuf = std::mem::replace(&mut self.recordbuf, BytesMut::new()).freeze();
-                    break;
+                    let recordbuf = std::mem::replace(&mut self.recordbuf, BytesMut::new());
+
+                    let recordbuf = recordbuf.freeze();
+                    let mut buf = recordbuf.clone();
+
+                    let xlogrec = XLogRecord::from_bytes(&mut buf);
+
+                    // XLOG_SWITCH records are special. If we see one, we need to skip
+                    // to the next WAL segment.
+                    if xlogrec.is_xlog_switch_record() {
+                        trace!("saw xlog switch record at {}", self.lsn);
+                        self.padlen =
+                            self.lsn.calc_padding(pg_constants::WAL_SEGMENT_SIZE as u64) as u32;
+                    } else {
+                        // Pad to an 8-byte boundary
+                        self.padlen = self.lsn.calc_padding(8u32) as u32;
+                    }
+
+                    let mut crc = crc32c_append(0, &recordbuf[XLOG_RECORD_CRC_OFFS + 4..]);
+                    crc = crc32c_append(crc, &recordbuf[0..XLOG_RECORD_CRC_OFFS]);
+                    if crc != xlogrec.xl_crc {
+                        return Err(WalDecodeError {
+                            msg: "WAL record crc mismatch".into(),
+                            lsn: self.lsn,
+                        });
+                    }
+
+                    // Always align resulting LSN on 0x8 boundary -- that is important for getPage()
+                    // and WalReceiver integration. Since this code is used both for WalReceiver and
+                    // initial WAL import let's force alignment right here.
+                    let result = (self.lsn.align(), recordbuf);
+                    return Ok(Some(result));
                }
                continue;
            }
        }
+        // check record boundaries

-        // We now have a record in the 'recordbuf' local variable.
-        let xlogrec = XLogRecord::from_slice(&recordbuf[0..XLOG_SIZE_OF_XLOG_RECORD]);
+        // deal with continuation records

-        let mut crc = 0;
-        crc = crc32c_append(crc, &recordbuf[XLOG_RECORD_CRC_OFFS + 4..]);
-        crc = crc32c_append(crc, &recordbuf[0..XLOG_RECORD_CRC_OFFS]);
-        if crc != xlogrec.xl_crc {
-            return Err(WalDecodeError {
-                msg: "WAL record crc mismatch".into(),
-                lsn: self.lsn,
-            });
-        }
-
-        // XLOG_SWITCH records are special. If we see one, we need to skip
-        // to the next WAL segment.
-        if xlogrec.is_xlog_switch_record() {
-            trace!("saw xlog switch record at {}", self.lsn);
-            self.padlen = self.lsn.calc_padding(pg_constants::WAL_SEGMENT_SIZE as u64) as u32;
-        } else {
-            // Pad to an 8-byte boundary
-            self.padlen = self.lsn.calc_padding(8u32) as u32;
-        }
-
-        // Always align resulting LSN on 0x8 boundary -- that is important for getPage()
-        // and WalReceiver integration. Since this code is used both for WalReceiver and
-        // initial WAL import let's force alignment right here.
-        let result = (self.lsn.align(), recordbuf);
-        Ok(Some(result))
+        // deal with xlog_switch records
    }
 }

--- a/pageserver/src/walreceiver.rs
+++ b/pageserver/src/walreceiver.rs
@@ -12,6 +12,7 @@ use crate::waldecoder::*;
 use crate::PageServerConf;
 use anyhow::{bail, Error, Result};
 use lazy_static::lazy_static;
+use log::*;
 use postgres::fallible_iterator::FallibleIterator;
 use postgres::replication::ReplicationIter;
 use postgres::{Client, NoTls, SimpleQueryMessage, SimpleQueryRow};
@@ -24,10 +25,8 @@ use std::str::FromStr;
 use std::sync::Mutex;
 use std::thread;
 use std::thread::sleep;
-use std::thread::JoinHandle;
 use std::thread_local;
 use std::time::{Duration, SystemTime};
-use tracing::*;
 use zenith_utils::lsn::Lsn;
 use zenith_utils::zid::ZTenantId;
 use zenith_utils::zid::ZTimelineId;
@@ -37,7 +36,6 @@ use zenith_utils::zid::ZTimelineId;
 //
 struct WalReceiverEntry {
    wal_producer_connstr: String,
-    wal_receiver_handle: Option<JoinHandle<()>>,
 }

 lazy_static! {
@@ -52,19 +50,6 @@ thread_local! {
    pub(crate) static IS_WAL_RECEIVER: Cell<bool> = Cell::new(false);
 }

-// Wait for walreceiver to stop
-// Now it stops when pageserver shutdown is requested.
-// In future we can make this more granular and send shutdown signals
-// per tenant/timeline to cancel inactive walreceivers.
-// TODO deal with blocking pg connections
-pub fn stop_wal_receiver(timelineid: ZTimelineId) {
-    let mut receivers = WAL_RECEIVERS.lock().unwrap();
-    if let Some(r) = receivers.get_mut(&timelineid) {
-        r.wal_receiver_handle.take();
-        // r.wal_receiver_handle.take().map(JoinHandle::join);
-    }
-}
-
 // Launch a new WAL receiver, or tell one that's running about change in connection string
 pub fn launch_wal_receiver(
    conf: &'static PageServerConf,
@@ -79,19 +64,19 @@ pub fn launch_wal_receiver(
            receiver.wal_producer_connstr = wal_producer_connstr.into();
        }
        None => {
-            let wal_receiver_handle = thread::Builder::new()
+            let receiver = WalReceiverEntry {
+                wal_producer_connstr: wal_producer_connstr.into(),
+            };
+            receivers.insert(timelineid, receiver);
+
+            // Also launch a new thread to handle this connection
+            let _walreceiver_thread = thread::Builder::new()
                .name("WAL receiver thread".into())
                .spawn(move || {
                    IS_WAL_RECEIVER.with(|c| c.set(true));
                    thread_main(conf, timelineid, tenantid);
                })
                .unwrap();
-
-            let receiver = WalReceiverEntry {
-                wal_producer_connstr: wal_producer_connstr.into(),
-                wal_receiver_handle: Some(wal_receiver_handle),
-            };
-            receivers.insert(timelineid, receiver);
        }
    };
 }
@@ -111,14 +96,16 @@ fn get_wal_producer_connstr(timelineid: ZTimelineId) -> String {
 // This is the entry point for the WAL receiver thread.
 //
 fn thread_main(conf: &'static PageServerConf, timelineid: ZTimelineId, tenantid: ZTenantId) {
-    let _enter = info_span!("WAL receiver", timeline = %timelineid, tenant = %tenantid).entered();
-    info!("WAL receiver thread started");
+    info!(
+        "WAL receiver thread started for timeline : '{}'",
+        timelineid
+    );

    //
    // Make a connection to the WAL safekeeper, or directly to the primary PostgreSQL server,
    // and start streaming WAL from it. If the connection is lost, keep retrying.
    //
-    while !tenant_mgr::shutdown_requested() {
+    loop {
        // Look up the current WAL producer address
        let wal_producer_connstr = get_wal_producer_connstr(timelineid);

@@ -132,7 +119,6 @@ fn thread_main(conf: &'static PageServerConf, timelineid: ZTimelineId, tenantid:
            sleep(Duration::from_secs(1));
        }
    }
-    debug!("WAL streaming shut down");
 }

 fn walreceiver_main(
@@ -183,8 +169,8 @@ fn walreceiver_main(
    startpoint += startpoint.calc_padding(8u32);

    info!(
-        "last_record_lsn {} starting replication from {}, server is at {}...",
-        last_rec_lsn, startpoint, end_of_wal
+        "last_record_lsn {} starting replication from {} for timeline {}, server is at {}...",
+        last_rec_lsn, startpoint, timelineid, end_of_wal
    );

    let query = format!("START_REPLICATION PHYSICAL {}", startpoint);
@@ -212,32 +198,27 @@ fn walreceiver_main(
                waldecoder.feed_bytes(data);

                while let Some((lsn, recdata)) = waldecoder.poll_decode()? {
-                    let _enter = info_span!("processing record", lsn = %lsn).entered();
+                    // Save old checkpoint value to compare with it after decoding WAL record
+                    let old_checkpoint_bytes = checkpoint.encode();
+                    let decoded = decode_wal_record(recdata.clone());

                    // It is important to deal with the aligned records as lsn in getPage@LSN is
                    // aligned and can be several bytes bigger. Without this alignment we are
                    // at risk of hittind a deadlock.
                    assert!(lsn.is_aligned());

-                    let writer = timeline.writer();
-
-                    let mut checkpoint_modified = false;
-
-                    let decoded = decode_wal_record(recdata.clone());
                    restore_local_repo::save_decoded_record(
                        &mut checkpoint,
-                        &mut checkpoint_modified,
-                        writer.as_ref(),
+                        &*timeline,
                        &decoded,
                        recdata,
                        lsn,
                    )?;

+                    let new_checkpoint_bytes = checkpoint.encode();
                    // Check if checkpoint data was updated by save_decoded_record
-                    if checkpoint_modified {
-                        let new_checkpoint_bytes = checkpoint.encode();
-
-                        writer.put_page_image(
+                    if new_checkpoint_bytes != old_checkpoint_bytes {
+                        timeline.put_page_image(
                            RelishTag::Checkpoint,
                            0,
                            lsn,
@@ -247,7 +228,7 @@ fn walreceiver_main(

                    // Now that this record has been fully handled, including updating the
                    // checkpoint data, let the repository know that it is up-to-date to this LSN
-                    writer.advance_last_record_lsn(lsn);
+                    timeline.advance_last_record_lsn(lsn);
                    last_rec_lsn = lsn;
                }

@@ -292,11 +273,6 @@ fn walreceiver_main(

            physical_stream.standby_status_update(write_lsn, flush_lsn, apply_lsn, ts, NO_REPLY)?;
        }
-
-        if tenant_mgr::shutdown_requested() {
-            debug!("stop walreceiver because pageserver shutdown is requested");
-            break;
-        }
    }
    Ok(())
 }
--- a/pageserver/src/walredo.rs
+++ b/pageserver/src/walredo.rs
@@ -82,7 +82,7 @@ pub trait WalRedoManager: Send + Sync {
        blknum: u32,
        lsn: Lsn,
        base_img: Option<Bytes>,
-        records: Vec<(Lsn, WALRecord)>,
+        records: Vec<WALRecord>,
    ) -> Result<Bytes, WalRedoError>;
 }

@@ -99,7 +99,7 @@ impl crate::walredo::WalRedoManager for DummyRedoManager {
        _blknum: u32,
        _lsn: Lsn,
        _base_img: Option<Bytes>,
-        _records: Vec<(Lsn, WALRecord)>,
+        _records: Vec<WALRecord>,
    ) -> Result<Bytes, WalRedoError> {
        Err(WalRedoError::InvalidState)
    }
@@ -150,7 +150,7 @@ struct WalRedoRequest {
    lsn: Lsn,

    base_img: Option<Bytes>,
-    records: Vec<(Lsn, WALRecord)>,
+    records: Vec<WALRecord>,
 }

 /// An error happened in WAL redo
@@ -179,7 +179,7 @@ impl WalRedoManager for PostgresRedoManager {
        blknum: u32,
        lsn: Lsn,
        base_img: Option<Bytes>,
-        records: Vec<(Lsn, WALRecord)>,
+        records: Vec<WALRecord>,
    ) -> Result<Bytes, WalRedoError> {
        let start_time;
        let lock_time;
@@ -277,7 +277,7 @@ impl PostgresRedoManager {
                page.extend_from_slice(&ZERO_PAGE);
            }
            // Apply all collected WAL records
-            for (_lsn, record) in records {
+            for record in records {
                let mut buf = record.rec.clone();

                WAL_REDO_RECORD_COUNTER.inc();
@@ -544,7 +544,7 @@ impl PostgresRedoProcess {
        &mut self,
        tag: BufferTag,
        base_img: Option<Bytes>,
-        records: &[(Lsn, WALRecord)],
+        records: &[WALRecord],
    ) -> Result<Bytes, std::io::Error> {
        let stdout = &mut self.stdout;
        // Buffer the writes to avoid a lot of small syscalls.
@@ -565,16 +565,22 @@ impl PostgresRedoProcess {
                stdin.write_all(&build_begin_redo_for_block_msg(tag)),
            )
            .await??;
-            if let Some(img) = base_img {
-                timeout(TIMEOUT, stdin.write_all(&build_push_page_msg(tag, &img))).await??;
+            if base_img.is_some() {
+                timeout(
+                    TIMEOUT,
+                    stdin.write_all(&build_push_page_msg(tag, base_img.unwrap())),
+                )
+                .await??;
            }

            // Send WAL records.
-            for (lsn, rec) in records.iter() {
+            for rec in records.iter() {
+                let r = rec.clone();
+
                WAL_REDO_RECORD_COUNTER.inc();

                stdin
-                    .write_all(&build_apply_record_msg(*lsn, &rec.rec))
+                    .write_all(&build_apply_record_msg(r.lsn, r.rec))
                    .await?;

                //debug!("sent WAL record to wal redo postgres process ({:X}/{:X}",
@@ -611,41 +617,58 @@ impl PostgresRedoProcess {
 // process. See vendor/postgres/src/backend/tcop/zenith_wal_redo.c for
 // explanation of the protocol.

-fn build_begin_redo_for_block_msg(tag: BufferTag) -> Vec<u8> {
+fn build_begin_redo_for_block_msg(tag: BufferTag) -> Bytes {
    let len = 4 + 1 + 4 * 4;
-    let mut buf = Vec::with_capacity(1 + len);
+    let mut buf = BytesMut::with_capacity(1 + len);

    buf.put_u8(b'B');
    buf.put_u32(len as u32);

-    tag.ser_into(&mut buf)
+    // FIXME: this is a temporary hack that should go away when we refactor
+    // the postgres protocol serialization + handlers.
+    //
+    // BytesMut is a dynamic growable buffer, used a lot in tokio code but
+    // not in the std library. To write to a BytesMut from a serde serializer,
+    // we need to either:
+    // - pre-allocate the required buffer space. This is annoying because we
+    //   shouldn't care what the exact serialized size is-- that's the
+    //   serializer's job.
+    // - Or, we need to create a temporary "writer" (which implements the
+    //   `Write` trait). It's a bit awkward, because the writer consumes the
+    //   underlying BytesMut, and we need to extract it later with
+    //   `into_inner`.
+    let mut writer = buf.writer();
+    tag.ser_into(&mut writer)
        .expect("serialize BufferTag should always succeed");
+    let buf = writer.into_inner();

    debug_assert!(buf.len() == 1 + len);

-    buf
+    buf.freeze()
 }

-fn build_push_page_msg(tag: BufferTag, base_img: &[u8]) -> Vec<u8> {
+fn build_push_page_msg(tag: BufferTag, base_img: Bytes) -> Bytes {
    assert!(base_img.len() == 8192);

    let len = 4 + 1 + 4 * 4 + base_img.len();
-    let mut buf = Vec::with_capacity(1 + len);
+    let mut buf = BytesMut::with_capacity(1 + len);

    buf.put_u8(b'P');
    buf.put_u32(len as u32);
-    tag.ser_into(&mut buf)
+    let mut writer = buf.writer();
+    tag.ser_into(&mut writer)
        .expect("serialize BufferTag should always succeed");
+    let mut buf = writer.into_inner();
    buf.put(base_img);

    debug_assert!(buf.len() == 1 + len);

-    buf
+    buf.freeze()
 }

-fn build_apply_record_msg(endlsn: Lsn, rec: &[u8]) -> Vec<u8> {
+fn build_apply_record_msg(endlsn: Lsn, rec: Bytes) -> Bytes {
    let len = 4 + 8 + rec.len();
-    let mut buf: Vec<u8> = Vec::with_capacity(1 + len);
+    let mut buf = BytesMut::with_capacity(1 + len);

    buf.put_u8(b'A');
    buf.put_u32(len as u32);
@@ -654,19 +677,21 @@ fn build_apply_record_msg(endlsn: Lsn, rec: &[u8]) -> Vec<u8> {

    debug_assert!(buf.len() == 1 + len);

-    buf
+    buf.freeze()
 }

-fn build_get_page_msg(tag: BufferTag) -> Vec<u8> {
+fn build_get_page_msg(tag: BufferTag) -> Bytes {
    let len = 4 + 1 + 4 * 4;
-    let mut buf = Vec::with_capacity(1 + len);
+    let mut buf = BytesMut::with_capacity(1 + len);

    buf.put_u8(b'G');
    buf.put_u32(len as u32);
-    tag.ser_into(&mut buf)
+    let mut writer = buf.writer();
+    tag.ser_into(&mut writer)
        .expect("serialize BufferTag should always succeed");
+    let buf = writer.into_inner();

    debug_assert!(buf.len() == 1 + len);

-    buf
+    buf.freeze()
 }
--- a/postgres_ffi/src/xlog_utils.rs
+++ b/postgres_ffi/src/xlog_utils.rs
@@ -329,12 +329,7 @@ pub fn main() {
 }

 impl XLogRecord {
-    pub fn from_slice(buf: &[u8]) -> XLogRecord {
-        use zenith_utils::bin_ser::LeSer;
-        XLogRecord::des(buf).unwrap()
-    }
-
-    pub fn from_bytes<B: Buf>(buf: &mut B) -> XLogRecord {
+    pub fn from_bytes(buf: &mut Bytes) -> XLogRecord {
        use zenith_utils::bin_ser::LeSer;
        XLogRecord::des_from(&mut buf.reader()).unwrap()
    }
@@ -382,12 +377,10 @@ impl CheckPoint {
        Ok(CheckPoint::des(buf)?)
    }

-    /// Update next XID based on provided new_xid and stored epoch.
-    /// Next XID should be greater than new_xid. This handles 32-bit
-    /// XID wraparound correctly.
-    ///
-    /// Returns 'true' if the XID was updated.
-    pub fn update_next_xid(&mut self, xid: u32) -> bool {
+    // Update next XID based on provided new_xid and stored epoch.
+    // Next XID should be greater than new_xid.
+    // Also take in account 32-bit wrap-around.
+    pub fn update_next_xid(&mut self, xid: u32) {
        let xid = xid.wrapping_add(XID_CHECKPOINT_INTERVAL - 1) & !(XID_CHECKPOINT_INTERVAL - 1);
        let full_xid = self.nextXid.value;
        let new_xid = std::cmp::max(xid + 1, pg_constants::FIRST_NORMAL_TRANSACTION_ID);
@@ -398,14 +391,10 @@ impl CheckPoint {
                // wrap-around
                epoch += 1;
            }
-            let nextXid = (epoch << 32) | new_xid as u64;
-
-            if nextXid != self.nextXid.value {
-                self.nextXid = FullTransactionId { value: nextXid };
-                return true;
-            }
+            self.nextXid = FullTransactionId {
+                value: (epoch << 32) | new_xid as u64,
+            };
        }
-        false
    }
 }

--- a/proxy/src/main.rs
+++ b/proxy/src/main.rs
@@ -99,7 +99,7 @@ fn main() -> anyhow::Result<()> {
                .long("uri")
                .takes_value(true)
                .help("redirect unauthenticated users to given uri")
-                .default_value("http://127.0.0.1:3000/psql_session/"),
+                .default_value("http://localhost:3000/psql_session/"),
        )
        .arg(
            Arg::with_name("auth-endpoint")
@@ -107,7 +107,7 @@ fn main() -> anyhow::Result<()> {
                .long("auth-endpoint")
                .takes_value(true)
                .help("redirect unauthenticated users to given uri")
-                .default_value("http://127.0.0.1:3000/authenticate_proxy_request/"),
+                .default_value("http://localhost:3000/authenticate_proxy_request/"),
        )
        .arg(
            Arg::with_name("ssl-key")
--- a/proxy/src/mgmt.rs
+++ b/proxy/src/mgmt.rs
@@ -34,7 +34,7 @@ pub fn thread_main(state: &'static ProxyState, listener: TcpListener) -> anyhow:

 pub fn mgmt_conn_main(state: &'static ProxyState, socket: TcpStream) -> anyhow::Result<()> {
    let mut conn_handler = MgmtHandler { state };
-    let pgbackend = PostgresBackend::new(socket, AuthType::Trust, None, true)?;
+    let pgbackend = PostgresBackend::new(socket, AuthType::Trust, None)?;
    pgbackend.run(&mut conn_handler)
 }

--- a/proxy/src/proxy.rs
+++ b/proxy/src/proxy.rs
@@ -64,7 +64,6 @@ pub fn proxy_conn_main(
            socket,
            postgres_backend::AuthType::MD5,
            state.conf.ssl_config.clone(),
-            false,
        )?,
        md5_salt: [0u8; 4],
        psql_session_id: "".into(),
--- a/test_runner/batch_others/test_branch_behind.py
+++ b/test_runner/batch_others/test_branch_behind.py
@@ -1,6 +1,6 @@
 import subprocess
 from fixtures.zenith_fixtures import PostgresFactory, ZenithPageserver
-from fixtures.log_helper import log
+

 pytest_plugins = ("fixtures.zenith_fixtures")

@@ -13,7 +13,7 @@ def test_branch_behind(zenith_cli, pageserver: ZenithPageserver, postgres: Postg
    zenith_cli.run(["branch", "test_branch_behind", "empty"])

    pgmain = postgres.create_start('test_branch_behind')
-    log.info("postgres is running on 'test_branch_behind' branch")
+    print("postgres is running on 'test_branch_behind' branch")

    main_pg_conn = pgmain.connect()
    main_cur = main_pg_conn.cursor()
@@ -27,7 +27,7 @@ def test_branch_behind(zenith_cli, pageserver: ZenithPageserver, postgres: Postg
    ''')
    main_cur.execute('SELECT pg_current_wal_insert_lsn()')
    lsn_a = main_cur.fetchone()[0]
-    log.info(f'LSN after 100 rows: {lsn_a}')
+    print('LSN after 100 rows: ' + lsn_a)

    # Insert some more rows. (This generates enough WAL to fill a few segments.)
    main_cur.execute('''
@@ -37,7 +37,7 @@ def test_branch_behind(zenith_cli, pageserver: ZenithPageserver, postgres: Postg
    ''')
    main_cur.execute('SELECT pg_current_wal_insert_lsn()')
    lsn_b = main_cur.fetchone()[0]
-    log.info(f'LSN after 200100 rows: {lsn_b}')
+    print('LSN after 200100 rows: ' + lsn_b)

    # Branch at the point where only 100 rows were inserted
    zenith_cli.run(["branch", "test_branch_behind_hundred", "test_branch_behind@" + lsn_a])
@@ -52,7 +52,7 @@ def test_branch_behind(zenith_cli, pageserver: ZenithPageserver, postgres: Postg

    main_cur.execute('SELECT pg_current_wal_insert_lsn()')
    lsn_c = main_cur.fetchone()[0]
-    log.info(f'LSN after 400100 rows: {lsn_c}')
+    print('LSN after 400100 rows: ' + lsn_c)

    # Branch at the point where only 200100 rows were inserted
    zenith_cli.run(["branch", "test_branch_behind_more", "test_branch_behind@" + lsn_b])
@@ -89,4 +89,4 @@ def test_branch_behind(zenith_cli, pageserver: ZenithPageserver, postgres: Postg
    try:
        zenith_cli.run(["branch", "test_branch_preinitdb", "test_branch_behind@0/42"])
    except subprocess.CalledProcessError:
-        log.info("Branch creation with pre-initdb LSN failed (as expected)")
+        print("Branch creation with pre-initdb LSN failed (as expected)")
--- a/test_runner/batch_others/test_clog_truncate.py
+++ b/test_runner/batch_others/test_clog_truncate.py
@@ -4,7 +4,6 @@ import os
 from contextlib import closing

 from fixtures.zenith_fixtures import PostgresFactory, ZenithPageserver
-from fixtures.log_helper import log

 pytest_plugins = ("fixtures.zenith_fixtures")

@@ -25,7 +24,7 @@ def test_clog_truncate(zenith_cli, pageserver: ZenithPageserver, postgres: Postg
    ]

    pg = postgres.create_start('test_clog_truncate', config_lines=config)
-    log.info('postgres is running on test_clog_truncate branch')
+    print('postgres is running on test_clog_truncate branch')

    # Install extension containing function needed for test
    pg.safe_psql('CREATE EXTENSION zenith_test_utils')
@@ -34,22 +33,22 @@ def test_clog_truncate(zenith_cli, pageserver: ZenithPageserver, postgres: Postg
    with closing(pg.connect()) as conn:
        with conn.cursor() as cur:
            cur.execute('select test_consume_xids(1000*1000*10);')
-            log.info('xids consumed')
+            print('xids consumed')

            # call a checkpoint to trigger TruncateSubtrans
            cur.execute('CHECKPOINT;')

            # ensure WAL flush
            cur.execute('select txid_current()')
-            log.info(cur.fetchone())
+            print(cur.fetchone())

    # wait for autovacuum to truncate the pg_xact
    # XXX Is it worth to add a timeout here?
    pg_xact_0000_path = os.path.join(pg.pg_xact_dir_path(), '0000')
-    log.info(f"pg_xact_0000_path = {pg_xact_0000_path}")
+    print("pg_xact_0000_path = " + pg_xact_0000_path)

    while os.path.isfile(pg_xact_0000_path):
-        log.info(f"file exists. wait for truncation. " "pg_xact_0000_path = {pg_xact_0000_path}")
+        print("file exists. wait for truncation. " "pg_xact_0000_path = " + pg_xact_0000_path)
        time.sleep(5)

    # checkpoint to advance latest lsn
@@ -60,14 +59,14 @@ def test_clog_truncate(zenith_cli, pageserver: ZenithPageserver, postgres: Postg
            lsn_after_truncation = cur.fetchone()[0]

    # create new branch after clog truncation and start a compute node on it
-    log.info(f'create branch at lsn_after_truncation {lsn_after_truncation}')
+    print('create branch at lsn_after_truncation ' + lsn_after_truncation)
    zenith_cli.run(
        ["branch", "test_clog_truncate_new", "test_clog_truncate@" + lsn_after_truncation])

    pg2 = postgres.create_start('test_clog_truncate_new')
-    log.info('postgres is running on test_clog_truncate_new branch')
+    print('postgres is running on test_clog_truncate_new branch')

    # check that new node doesn't contain truncated segment
    pg_xact_0000_path_new = os.path.join(pg2.pg_xact_dir_path(), '0000')
-    log.info(f"pg_xact_0000_path_new = {pg_xact_0000_path_new}")
+    print("pg_xact_0000_path_new = " + pg_xact_0000_path_new)
    assert os.path.isfile(pg_xact_0000_path_new) is False
--- a/test_runner/batch_others/test_config.py
+++ b/test_runner/batch_others/test_config.py
@@ -1,7 +1,6 @@
 from contextlib import closing

 from fixtures.zenith_fixtures import PostgresFactory, ZenithPageserver
-from fixtures.log_helper import log

 pytest_plugins = ("fixtures.zenith_fixtures")

@@ -15,7 +14,7 @@ def test_config(zenith_cli, pageserver: ZenithPageserver, postgres: PostgresFact

    # change config
    pg = postgres.create_start('test_config', config_lines=['log_min_messages=debug1'])
-    log.info('postgres is running on test_config branch')
+    print('postgres is running on test_config branch')

    with closing(pg.connect()) as conn:
        with conn.cursor() as cur:
--- a/test_runner/batch_others/test_createdropdb.py
+++ b/test_runner/batch_others/test_createdropdb.py
@@ -3,7 +3,6 @@ import pathlib

 from contextlib import closing
 from fixtures.zenith_fixtures import ZenithPageserver, PostgresFactory, ZenithCli, check_restored_datadir_content
-from fixtures.log_helper import log

 pytest_plugins = ("fixtures.zenith_fixtures")

@@ -20,7 +19,7 @@ def test_createdb(
    zenith_cli.run(["branch", "test_createdb", "empty"])

    pg = postgres.create_start('test_createdb')
-    log.info("postgres is running on 'test_createdb' branch")
+    print("postgres is running on 'test_createdb' branch")

    with closing(pg.connect()) as conn:
        with conn.cursor() as cur:
@@ -54,7 +53,7 @@ def test_dropdb(
    zenith_cli.run(["branch", "test_dropdb", "empty"])

    pg = postgres.create_start('test_dropdb')
-    log.info("postgres is running on 'test_dropdb' branch")
+    print("postgres is running on 'test_dropdb' branch")

    with closing(pg.connect()) as conn:
        with conn.cursor() as cur:
@@ -89,13 +88,13 @@ def test_dropdb(

    # Test that database subdir exists on the branch before drop
    dbpath = pathlib.Path(pg_before.pgdata_dir) / 'base' / str(dboid)
-    log.info(dbpath)
+    print(dbpath)

    assert os.path.isdir(dbpath) == True

    # Test that database subdir doesn't exist on the branch after drop
    dbpath = pathlib.Path(pg_after.pgdata_dir) / 'base' / str(dboid)
-    log.info(dbpath)
+    print(dbpath)

    assert os.path.isdir(dbpath) == False

--- a/test_runner/batch_others/test_createuser.py
+++ b/test_runner/batch_others/test_createuser.py
@@ -1,7 +1,6 @@
 from contextlib import closing

 from fixtures.zenith_fixtures import PostgresFactory, ZenithPageserver
-from fixtures.log_helper import log

 pytest_plugins = ("fixtures.zenith_fixtures")

@@ -13,7 +12,7 @@ def test_createuser(zenith_cli, pageserver: ZenithPageserver, postgres: Postgres
    zenith_cli.run(["branch", "test_createuser", "empty"])

    pg = postgres.create_start('test_createuser')
-    log.info("postgres is running on 'test_createuser' branch")
+    print("postgres is running on 'test_createuser' branch")

    with closing(pg.connect()) as conn:
        with conn.cursor() as cur:
--- a/test_runner/batch_others/test_multixact.py
+++ b/test_runner/batch_others/test_multixact.py
@@ -1,5 +1,4 @@
 from fixtures.zenith_fixtures import PostgresFactory, ZenithPageserver, check_restored_datadir_content
-from fixtures.log_helper import log

 pytest_plugins = ("fixtures.zenith_fixtures")

@@ -16,7 +15,7 @@ def test_multixact(pageserver: ZenithPageserver, postgres: PostgresFactory,
    zenith_cli.run(["branch", "test_multixact", "empty"])
    pg = postgres.create_start('test_multixact')

-    log.info("postgres is running on 'test_multixact' branch")
+    print("postgres is running on 'test_multixact' branch")
    pg_conn = pg.connect()
    cur = pg_conn.cursor()

@@ -56,7 +55,7 @@ def test_multixact(pageserver: ZenithPageserver, postgres: PostgresFactory,
    zenith_cli.run(["branch", "test_multixact_new", "test_multixact@" + lsn])
    pg_new = postgres.create_start('test_multixact_new')

-    log.info("postgres is running on 'test_multixact_new' branch")
+    print("postgres is running on 'test_multixact_new' branch")
    pg_new_conn = pg_new.connect()
    cur_new = pg_new_conn.cursor()

--- a/test_runner/batch_others/test_old_request_lsn.py
+++ b/test_runner/batch_others/test_old_request_lsn.py
@@ -1,7 +1,6 @@
 from contextlib import closing

 from fixtures.zenith_fixtures import PostgresFactory, ZenithPageserver
-from fixtures.log_helper import log

 pytest_plugins = ("fixtures.zenith_fixtures")

@@ -19,7 +18,7 @@ def test_old_request_lsn(zenith_cli, pageserver: ZenithPageserver, postgres: Pos
    # Create a branch for us
    zenith_cli.run(["branch", "test_old_request_lsn", "empty"])
    pg = postgres.create_start('test_old_request_lsn')
-    log.info('postgres is running on test_old_request_lsn branch')
+    print('postgres is running on test_old_request_lsn branch')

    pg_conn = pg.connect()
    cur = pg_conn.cursor()
@@ -47,7 +46,7 @@ def test_old_request_lsn(zenith_cli, pageserver: ZenithPageserver, postgres: Pos
        from pg_settings where name = 'shared_buffers'
    ''')
    row = cur.fetchone()
-    log.info(f'shared_buffers is {row[0]}, table size {row[1]}');
+    print(f'shared_buffers is {row[0]}, table size {row[1]}');
    assert int(row[0]) < int(row[1])

    cur.execute('VACUUM foo');
--- a/test_runner/batch_others/test_pageserver_restart.py
+++ b/test_runner/batch_others/test_pageserver_restart.py
@@ -5,7 +5,6 @@ import time
 from contextlib import closing
 from multiprocessing import Process, Value
 from fixtures.zenith_fixtures import WalAcceptorFactory, ZenithPageserver, PostgresFactory
-from fixtures.log_helper import log

 pytest_plugins = ("fixtures.zenith_fixtures")

@@ -41,7 +40,7 @@ def test_pageserver_restart(zenith_cli, pageserver: ZenithPageserver, postgres:
        from pg_settings where name = 'shared_buffers'
    ''')
    row = cur.fetchone()
-    log.info(f"shared_buffers is {row[0]}, table size {row[1]}");
+    print("shared_buffers is {}, table size {}", row[0], row[1]);
    assert int(row[0]) < int(row[1])

    # Stop and restart pageserver. This is a more or less graceful shutdown, although
--- a/test_runner/batch_others/test_pgbench.py
+++ b/test_runner/batch_others/test_pgbench.py
@@ -1,5 +1,4 @@
 from fixtures.zenith_fixtures import PostgresFactory
-from fixtures.log_helper import log

 pytest_plugins = ("fixtures.zenith_fixtures")

@@ -9,7 +8,7 @@ def test_pgbench(postgres: PostgresFactory, pg_bin, zenith_cli):
    zenith_cli.run(["branch", "test_pgbench", "empty"])

    pg = postgres.create_start('test_pgbench')
-    log.info("postgres is running on 'test_pgbench' branch")
+    print("postgres is running on 'test_pgbench' branch")

    connstr = pg.connstr()

--- a/test_runner/batch_others/test_restart_compute.py
+++ b/test_runner/batch_others/test_restart_compute.py
@@ -2,7 +2,6 @@ import pytest

 from contextlib import closing
 from fixtures.zenith_fixtures import ZenithPageserver, PostgresFactory
-from fixtures.log_helper import log

 pytest_plugins = ("fixtures.zenith_fixtures")

@@ -28,7 +27,7 @@ def test_restart_compute(

    pg = postgres.create_start('test_restart_compute',
                               wal_acceptors=wal_acceptor_connstrs)
-    log.info("postgres is running on 'test_restart_compute' branch")
+    print("postgres is running on 'test_restart_compute' branch")

    with closing(pg.connect()) as conn:
        with conn.cursor() as cur:
@@ -37,7 +36,7 @@ def test_restart_compute(
            cur.execute('SELECT sum(key) FROM t')
            r = cur.fetchone()
            assert r == (5000050000, )
-            log.info(f"res = {r}")
+            print("res = ", r)

    # Remove data directory and restart
    pg.stop_and_destroy().create_start('test_restart_compute',
@@ -50,7 +49,7 @@ def test_restart_compute(
            cur.execute('SELECT sum(key) FROM t')
            r = cur.fetchone()
            assert r == (5000050000, )
-            log.info(f"res = {r}")
+            print("res = ", r)

            # Insert another row
            cur.execute("INSERT INTO t VALUES (100001, 'payload2')")
@@ -58,7 +57,7 @@ def test_restart_compute(

            r = cur.fetchone()
            assert r == (100001, )
-            log.info(f"res = {r}")
+            print("res = ", r)

    # Again remove data directory and restart
    pg.stop_and_destroy().create_start('test_restart_compute',
@@ -73,7 +72,7 @@ def test_restart_compute(

            r = cur.fetchone()
            assert r == (100001, )
-            log.info(f"res = {r}")
+            print("res = ", r)

    # And again remove data directory and restart
    pg.stop_and_destroy().create_start('test_restart_compute',
@@ -86,4 +85,4 @@ def test_restart_compute(

            r = cur.fetchone()
            assert r == (100001, )
-            log.info(f"res = {r}")
+            print("res = ", r)
--- a/test_runner/batch_others/test_snapfiles_gc.py
+++ b/test_runner/batch_others/test_snapfiles_gc.py
@@ -1,14 +1,13 @@
 from contextlib import closing
 import psycopg2.extras
-import time
-from fixtures.log_helper import log
+import time;

 pytest_plugins = ("fixtures.zenith_fixtures")

 def print_gc_result(row):
-    log.info("GC duration {elapsed} ms".format_map(row));
-    log.info("  REL    total: {layer_relfiles_total}, needed_by_cutoff {layer_relfiles_needed_by_cutoff}, needed_by_branches: {layer_relfiles_needed_by_branches}, not_updated: {layer_relfiles_not_updated}, needed_as_tombstone {layer_relfiles_needed_as_tombstone}, removed: {layer_relfiles_removed}, dropped: {layer_relfiles_dropped}".format_map(row))
-    log.info("  NONREL total: {layer_nonrelfiles_total}, needed_by_cutoff {layer_nonrelfiles_needed_by_cutoff}, needed_by_branches: {layer_nonrelfiles_needed_by_branches}, not_updated: {layer_nonrelfiles_not_updated}, needed_as_tombstone {layer_nonrelfiles_needed_as_tombstone}, removed: {layer_nonrelfiles_removed}, dropped: {layer_nonrelfiles_dropped}".format_map(row))
+    print("GC duration {elapsed} ms".format_map(row));
+    print("  REL    total: {layer_relfiles_total}, needed_by_cutoff {layer_relfiles_needed_by_cutoff}, needed_by_branches: {layer_relfiles_needed_by_branches}, not_updated: {layer_relfiles_not_updated}, needed_as_tombstone {layer_relfiles_needed_as_tombstone}, removed: {layer_relfiles_removed}, dropped: {layer_relfiles_dropped}".format_map(row))
+    print("  NONREL total: {layer_nonrelfiles_total}, needed_by_cutoff {layer_nonrelfiles_needed_by_cutoff}, needed_by_branches: {layer_nonrelfiles_needed_by_branches}, not_updated: {layer_nonrelfiles_not_updated}, needed_as_tombstone {layer_nonrelfiles_needed_as_tombstone}, removed: {layer_nonrelfiles_removed}, dropped: {layer_nonrelfiles_dropped}".format_map(row))


 #
@@ -36,7 +35,7 @@ def test_layerfiles_gc(zenith_cli, pageserver, postgres, pg_bin):

                    cur.execute("select relfilenode from pg_class where oid = 'foo'::regclass");
                    row = cur.fetchone();
-                    log.info(f"relfilenode is {row[0]}");
+                    print("relfilenode is {}", row[0]);

                    # Run GC, to clear out any garbage left behind in the catalogs by
                    # the CREATE TABLE command. We want to have a clean slate with no garbage
@@ -51,7 +50,7 @@ def test_layerfiles_gc(zenith_cli, pageserver, postgres, pg_bin):
                    # update to confuse our numbers either.
                    cur.execute("DELETE FROM foo")

-                    log.info("Running GC before test")
+                    print("Running GC before test")
                    pscur.execute(f"do_gc {pageserver.initial_tenant} {timeline} 0")
                    row = pscur.fetchone()
                    print_gc_result(row);
@@ -62,7 +61,7 @@ def test_layerfiles_gc(zenith_cli, pageserver, postgres, pg_bin):
                    # Insert a row and run GC. Checkpoint should freeze the layer
                    # so that there is only the most recent image layer left for the rel,
                    # removing the old image and delta layer.
-                    log.info("Inserting one row and running GC")
+                    print("Inserting one row and running GC")
                    cur.execute("INSERT INTO foo VALUES (1)")
                    pscur.execute(f"do_gc {pageserver.initial_tenant} {timeline} 0")
                    row = pscur.fetchone()
@@ -74,7 +73,7 @@ def test_layerfiles_gc(zenith_cli, pageserver, postgres, pg_bin):
                    # Insert two more rows and run GC.
                    # This should create new image and delta layer file with the new contents, and
                    # then remove the old one image and the just-created delta layer.
-                    log.info("Inserting two more rows and running GC")
+                    print("Inserting two more rows and running GC")
                    cur.execute("INSERT INTO foo VALUES (2)")
                    cur.execute("INSERT INTO foo VALUES (3)")

@@ -86,7 +85,7 @@ def test_layerfiles_gc(zenith_cli, pageserver, postgres, pg_bin):
                    assert row['layer_relfiles_dropped'] == 0

                    # Do it again. Should again create two new layer files and remove old ones.
-                    log.info("Inserting two more rows and running GC")
+                    print("Inserting two more rows and running GC")
                    cur.execute("INSERT INTO foo VALUES (2)")
                    cur.execute("INSERT INTO foo VALUES (3)")

@@ -98,7 +97,7 @@ def test_layerfiles_gc(zenith_cli, pageserver, postgres, pg_bin):
                    assert row['layer_relfiles_dropped'] == 0

                    # Run GC again, with no changes in the database. Should not remove anything.
-                    log.info("Run GC again, with nothing to do")
+                    print("Run GC again, with nothing to do")
                    pscur.execute(f"do_gc {pageserver.initial_tenant} {timeline} 0")
                    row = pscur.fetchone()
                    print_gc_result(row);
@@ -109,7 +108,7 @@ def test_layerfiles_gc(zenith_cli, pageserver, postgres, pg_bin):
                    #
                    # Test DROP TABLE checks that relation data and metadata was deleted by GC from object storage
                    #
-                    log.info("Drop table and run GC again");
+                    print("Drop table and run GC again");
                    cur.execute("DROP TABLE foo")

                    pscur.execute(f"do_gc {pageserver.initial_tenant} {timeline} 0")
--- a/test_runner/batch_others/test_timeline_size.py
+++ b/test_runner/batch_others/test_timeline_size.py
@@ -2,7 +2,7 @@ from contextlib import closing
 from uuid import UUID
 import psycopg2.extras
 from fixtures.zenith_fixtures import PostgresFactory, ZenithPageserver
-from fixtures.log_helper import log
+

 def test_timeline_size(
    zenith_cli, pageserver: ZenithPageserver, postgres: PostgresFactory, pg_bin
@@ -15,7 +15,7 @@ def test_timeline_size(
    assert res["current_logical_size"] == res["current_logical_size_non_incremental"]

    pgmain = postgres.create_start("test_timeline_size")
-    log.info("postgres is running on 'test_timeline_size' branch")
+    print("postgres is running on 'test_timeline_size' branch")

    with closing(pgmain.connect()) as conn:
        with conn.cursor() as cur:
--- a/test_runner/batch_others/test_twophase.py
+++ b/test_runner/batch_others/test_twophase.py
@@ -1,7 +1,7 @@
 import os

 from fixtures.zenith_fixtures import PostgresFactory, ZenithPageserver, PgBin
-from fixtures.log_helper import log
+

 pytest_plugins = ("fixtures.zenith_fixtures")

@@ -13,7 +13,7 @@ def test_twophase(zenith_cli, pageserver: ZenithPageserver, postgres: PostgresFa
    zenith_cli.run(["branch", "test_twophase", "empty"])

    pg = postgres.create_start('test_twophase', config_lines=['max_prepared_transactions=5'])
-    log.info("postgres is running on 'test_twophase' branch")
+    print("postgres is running on 'test_twophase' branch")

    conn = pg.connect()
    cur = conn.cursor()
@@ -45,7 +45,7 @@ def test_twophase(zenith_cli, pageserver: ZenithPageserver, postgres: PostgresFa
    cur.execute('CHECKPOINT')

    twophase_files = os.listdir(pg.pg_twophase_dir_path())
-    log.info(twophase_files)
+    print(twophase_files)
    assert len(twophase_files) == 4

    cur.execute("COMMIT PREPARED 'insert_three'")
@@ -53,7 +53,7 @@ def test_twophase(zenith_cli, pageserver: ZenithPageserver, postgres: PostgresFa
    cur.execute('CHECKPOINT')

    twophase_files = os.listdir(pg.pg_twophase_dir_path())
-    log.info(twophase_files)
+    print(twophase_files)
    assert len(twophase_files) == 2

    # Create a branch with the transaction in prepared state
@@ -67,7 +67,7 @@ def test_twophase(zenith_cli, pageserver: ZenithPageserver, postgres: PostgresFa

    # Check that we restored only needed twophase files
    twophase_files2 = os.listdir(pg2.pg_twophase_dir_path())
-    log.info(twophase_files2)
+    print(twophase_files2)
    assert twophase_files2.sort() == twophase_files.sort()

    conn2 = pg2.connect()
--- a/test_runner/batch_others/test_vm_bits.py
+++ b/test_runner/batch_others/test_vm_bits.py
@@ -1,5 +1,4 @@
 from fixtures.zenith_fixtures import PostgresFactory, ZenithPageserver
-from fixtures.log_helper import log

 pytest_plugins = ("fixtures.zenith_fixtures")

@@ -12,7 +11,7 @@ def test_vm_bit_clear(pageserver: ZenithPageserver, postgres: PostgresFactory, p
    zenith_cli.run(["branch", "test_vm_bit_clear", "empty"])
    pg = postgres.create_start('test_vm_bit_clear')

-    log.info("postgres is running on 'test_vm_bit_clear' branch")
+    print("postgres is running on 'test_vm_bit_clear' branch")
    pg_conn = pg.connect()
    cur = pg_conn.cursor()

@@ -64,7 +63,7 @@ def test_vm_bit_clear(pageserver: ZenithPageserver, postgres: PostgresFactory, p
    # server at the right point-in-time avoids that full-page image.
    pg_new = postgres.create_start('test_vm_bit_clear_new')

-    log.info("postgres is running on 'test_vm_bit_clear_new' branch")
+    print("postgres is running on 'test_vm_bit_clear_new' branch")
    pg_new_conn = pg_new.connect()
    cur_new = pg_new_conn.cursor()

--- a/test_runner/batch_others/test_wal_acceptor.py
+++ b/test_runner/batch_others/test_wal_acceptor.py
@@ -9,7 +9,6 @@ from contextlib import closing
 from multiprocessing import Process, Value
 from fixtures.zenith_fixtures import WalAcceptorFactory, ZenithPageserver, PostgresFactory, PgBin
 from fixtures.utils import lsn_to_hex, mkdir_if_needed
-from fixtures.log_helper import log

 pytest_plugins = ("fixtures.zenith_fixtures")

@@ -204,7 +203,6 @@ def test_race_conditions(zenith_cli, pageserver: ZenithPageserver, postgres: Pos
    stop_value.value = 1
    proc.join()

-
 class ProposerPostgres:
    """Object for running safekeepers sync with walproposer"""
    def __init__(self, pgdata_dir: str, pg_bin: PgBin, timeline_id: str, tenant_id: str):
@@ -286,38 +284,10 @@ def test_sync_safekeepers(repo_dir: str, pg_bin: PgBin, wa_factory: WalAcceptorF
        )
        lsn_hex = lsn_to_hex(res["inserted_wal"]["end_lsn"])
        lsn_after_append.append(lsn_hex)
-        log.info(f"safekeeper[{i}] lsn after append: {lsn_hex}")
+        print(f"safekeeper[{i}] lsn after append: {lsn_hex}")

    # run sync safekeepers
    lsn_after_sync = pg.sync_safekeepers()
-    log.info(f"lsn after sync = {lsn_after_sync}")
+    print(f"lsn after sync = {lsn_after_sync}")

    assert all(lsn_after_sync == lsn for lsn in lsn_after_append)
-
-
-def test_timeline_status(zenith_cli, pageserver, postgres, wa_factory: WalAcceptorFactory):
-    wa_factory.start_n_new(1)
-
-    zenith_cli.run(["branch", "test_timeline_status", "empty"])
-    pg = postgres.create_start('test_timeline_status', wal_acceptors=wa_factory.get_connstrs())
-
-    wa = wa_factory.instances[0]
-    wa_http_cli = wa.http_client()
-    wa_http_cli.check_status()
-
-    # learn zenith timeline from compute
-    tenant_id = pg.safe_psql("show zenith.zenith_tenant")[0][0]
-    timeline_id = pg.safe_psql("show zenith.zenith_timeline")[0][0]
-
-    # fetch something sensible from status
-    epoch = wa_http_cli.timeline_status(tenant_id, timeline_id).acceptor_epoch
-
-    pg.safe_psql("create table t(i int)")
-
-    # ensure epoch goes up after reboot
-    pg.stop().start()
-    pg.safe_psql("insert into t values(10)")
-
-    epoch_after_reboot = wa_http_cli.timeline_status(tenant_id,
-                                                     timeline_id).acceptor_epoch
-    assert epoch_after_reboot > epoch
--- a/test_runner/batch_others/test_wal_acceptor_async.py
+++ b/test_runner/batch_others/test_wal_acceptor_async.py
@@ -3,10 +3,9 @@ import asyncpg
 import random

 from fixtures.zenith_fixtures import WalAcceptor, WalAcceptorFactory, ZenithPageserver, PostgresFactory, Postgres
-from fixtures.log_helper import getLogger
 from typing import List
+from fixtures.utils import debug_print

-log = getLogger('root.wal_acceptor_async')
 pytest_plugins = ("fixtures.zenith_fixtures")


@@ -64,18 +63,18 @@ class WorkerStats(object):
        self.counters[worker_id] += 1

    def check_progress(self):
-        log.debug("Workers progress: {}".format(self.counters))
+        debug_print("Workers progress: {}".format(self.counters))

        # every worker should finish at least one tx
        assert all(cnt > 0 for cnt in self.counters)

        progress = sum(self.counters)
-        log.info('All workers made {} transactions'.format(progress))
+        print('All workers made {} transactions'.format(progress))


 async def run_random_worker(stats: WorkerStats, pg: Postgres, worker_id, n_accounts, max_transfer):
    pg_conn = await pg.connect_async()
-    log.debug('Started worker {}'.format(worker_id))
+    debug_print('Started worker {}'.format(worker_id))

    while stats.running:
        from_uid = random.randint(0, n_accounts - 1)
@@ -85,9 +84,9 @@ async def run_random_worker(stats: WorkerStats, pg: Postgres, worker_id, n_accou
        await bank_transfer(pg_conn, from_uid, to_uid, amount)
        stats.inc_progress(worker_id)

-        log.debug('Executed transfer({}) {} => {}'.format(amount, from_uid, to_uid))
+        debug_print('Executed transfer({}) {} => {}'.format(amount, from_uid, to_uid))

-    log.debug('Finished worker {}'.format(worker_id))
+    debug_print('Finished worker {}'.format(worker_id))

    await pg_conn.close()

@@ -135,7 +134,7 @@ async def run_restarts_under_load(pg: Postgres, acceptors: List[WalAcceptor], n_

        victim.start()

-    log.info('Iterations are finished, exiting coroutines...')
+    print('Iterations are finished, exiting coroutines...')
    stats.running = False
    # await all workers
    await asyncio.gather(*workers)
--- a/test_runner/batch_pg_regress/test_zenith_regress.py
+++ b/test_runner/batch_pg_regress/test_zenith_regress.py
@@ -2,7 +2,6 @@ import os

 from fixtures.utils import mkdir_if_needed
 from fixtures.zenith_fixtures import PageserverPort, PostgresFactory, check_restored_datadir_content
-from fixtures.log_helper import log

 pytest_plugins = ("fixtures.zenith_fixtures")

@@ -39,7 +38,7 @@ def test_zenith_regress(postgres: PostgresFactory, pg_bin, zenith_cli, test_outp
        '--inputdir={}'.format(src_path),
    ]

-    log.info(pg_regress_command)
+    print(pg_regress_command)
    env = {
        'PGPORT': str(pg.port),
        'PGUSER': pg.username,
--- a/test_runner/fixtures/benchmark_fixture.py
+++ b/test_runner/fixtures/benchmark_fixture.py
@@ -1,3 +1,5 @@
+from pprint import pprint
+
 import os
 import re
 import timeit
--- a/test_runner/fixtures/log_helper.py
+++ b/test_runner/fixtures/log_helper.py
@@ -1,44 +0,0 @@
-import logging
-import logging.config
-
-"""
-This file configures logging to use in python tests.
-Logs are automatically captured and shown in their
-own section after all tests are executed. 
-
-To see logs for all (even successful) tests, run
-pytest with the following command:
- `pipenv run pytest -n8 -rA`
-
-Other log config can be set in pytest.ini file.
-You can add `log_cli = true` to it to watch
-logs in real time.
-
-To get more info about logging with pytest, see
-https://docs.pytest.org/en/6.2.x/logging.html
-"""
-
-# this config is only used for default log levels,
-# log format is specified in pytest.ini file
-LOGGING = {
-    "version": 1,
-    "loggers": {
-        "root": {
-            "level": "INFO"
-        },
-        "root.wal_acceptor_async": {
-            "level": "INFO" # a lot of logs on DEBUG level
-        }
-    }
-}
-
-def getLogger(name='root') -> logging.Logger:
-    """Method to get logger for tests.
-    
-    Should be used to get correctly initialized logger. """
-    return logging.getLogger(name)
-
-# default logger for tests
-log = getLogger()
-
-logging.config.dictConfig(LOGGING)
--- a/test_runner/fixtures/utils.py
+++ b/test_runner/fixtures/utils.py
@@ -2,7 +2,7 @@ import os
 import subprocess

 from typing import Any, List
-from fixtures.log_helper import log
+

 def get_self_dir() -> str:
    """ Get the path to the directory where this script lives. """
@@ -39,7 +39,7 @@ def subprocess_capture(capture_dir: str, cmd: List[str], **kwargs: Any) -> str:

    with open(stdout_filename, 'w') as stdout_f:
        with open(stderr_filename, 'w') as stderr_f:
-            log.info('(capturing output to "{}.stdout")'.format(base))
+            print('(capturing output to "{}.stdout")'.format(base))
            subprocess.run(cmd, **kwargs, stdout=stdout_f, stderr=stderr_f)

    return basepath
@@ -58,6 +58,14 @@ def global_counter() -> int:
    _global_counter += 1
    return _global_counter

+def debug_print(*args, **kwargs) -> None:
+    """ Print to the console if TEST_DEBUG_PRINT is set in env.
+    
+    All parameters are passed to print().
+    """
+    if os.environ.get('TEST_DEBUG_PRINT') is not None:
+        print(*args, **kwargs)
+
 def lsn_to_hex(num: int) -> str:
    """ Convert lsn from int to standard hex notation. """
    return "{:X}/{:X}".format(num >> 32, num & 0xffffffff)
--- a/test_runner/fixtures/zenith_fixtures.py
+++ b/test_runner/fixtures/zenith_fixtures.py
@@ -27,8 +27,6 @@ from typing_extensions import Literal
 import requests

 from .utils import (get_self_dir, mkdir_if_needed, subprocess_capture)
-from fixtures.log_helper import log
-
 """
 This file contains pytest fixtures. A fixture is a test resource that can be
 summoned by placing its name in the test's arguments.
@@ -66,7 +64,7 @@ def pytest_configure(config):
         raise Exception('Too many workers configured. Cannot distrubute ports for services.')

    # does not use -c as it is not supported on macOS
-    cmd = ['pgrep', 'pageserver|postgres|safekeeper']
+    cmd = ['pgrep', 'pageserver|postgres|wal_acceptor']
    result = subprocess.run(cmd, stdout=subprocess.DEVNULL)
    if result.returncode == 0:
        # returncode of 0 means it found something.
@@ -74,7 +72,7 @@ def pytest_configure(config):
        # result of the test.
        # NOTE this shows as an internal pytest error, there might be a better way
        raise Exception(
-            'Found interfering processes running. Stop all Zenith pageservers, nodes, safekeepers, as well as stand-alone Postgres.'
+            'Found interfering processes running. Stop all Zenith pageservers, nodes, WALs, as well as stand-alone Postgres.'
        )


@@ -190,13 +188,13 @@ class ZenithCli:

        >>> result = zenith_cli.run(...)
        >>> assert result.stderr == ""
-        >>> log.info(result.stdout)
+        >>> print(result.stdout)
        """

        assert type(arguments) == list

        args = [self.bin_zenith] + arguments
-        log.info('Running command "{}"'.format(' '.join(args)))
+        print('Running command "{}"'.format(' '.join(args)))

        # Interceipt CalledProcessError and print more info
        try:
@@ -213,7 +211,7 @@ class ZenithCli:
              stdout: {exc.stdout}
              stderr: {exc.stderr}
            """
-            log.info(msg)
+            print(msg)

            raise Exception(msg) from exc

@@ -235,16 +233,16 @@ class ZenithPageserverHttpClient(requests.Session):
            self.headers['Authorization'] = f'Bearer {auth_token}'

    def check_status(self):
-        self.get(f"http://127.0.0.1:{self.port}/v1/status").raise_for_status()
+        self.get(f"http://localhost:{self.port}/v1/status").raise_for_status()

    def branch_list(self, tenant_id: uuid.UUID) -> List[Dict]:
-        res = self.get(f"http://127.0.0.1:{self.port}/v1/branch/{tenant_id.hex}")
+        res = self.get(f"http://localhost:{self.port}/v1/branch/{tenant_id.hex}")
        res.raise_for_status()
        return res.json()

    def branch_create(self, tenant_id: uuid.UUID, name: str, start_point: str) -> Dict:
        res = self.post(
-            f"http://127.0.0.1:{self.port}/v1/branch",
+            f"http://localhost:{self.port}/v1/branch",
            json={
                'tenant_id': tenant_id.hex,
                'name': name,
@@ -256,19 +254,19 @@ class ZenithPageserverHttpClient(requests.Session):

    def branch_detail(self, tenant_id: uuid.UUID, name: str) -> Dict:
        res = self.get(
-            f"http://127.0.0.1:{self.port}/v1/branch/{tenant_id.hex}/{name}",
+            f"http://localhost:{self.port}/v1/branch/{tenant_id.hex}/{name}",
        )
        res.raise_for_status()
        return res.json()

    def tenant_list(self) -> List[str]:
-        res = self.get(f"http://127.0.0.1:{self.port}/v1/tenant")
+        res = self.get(f"http://localhost:{self.port}/v1/tenant")
        res.raise_for_status()
        return res.json()

    def tenant_create(self, tenant_id: uuid.UUID):
        res = self.post(
-            f"http://127.0.0.1:{self.port}/v1/tenant",
+            f"http://localhost:{self.port}/v1/tenant",
            json={
                'tenant_id': tenant_id.hex,
            },
@@ -277,7 +275,7 @@ class ZenithPageserverHttpClient(requests.Session):
        return res.json()

    def get_metrics(self) -> str:
-        res = self.get(f"http://127.0.0.1:{self.port}/metrics")
+        res = self.get(f"http://localhost:{self.port}/metrics")
        res.raise_for_status()
        return res.text

@@ -348,7 +346,7 @@ class PageserverPort:
 class ZenithPageserver(PgProtocol):
    """ An object representing a running pageserver. """
    def __init__(self, zenith_cli: ZenithCli, repo_dir: str, port: PageserverPort):
-        super().__init__(host='127.0.0.1', port=port.pg)
+        super().__init__(host='localhost', port=port.pg)
        self.zenith_cli = zenith_cli
        self.running = False
        self.initial_tenant = None
@@ -377,7 +375,6 @@ class ZenithPageserver(PgProtocol):
        Start the page server.
        Returns self.
        """
-        assert self.running == False

        self.zenith_cli.run(['start'])
        self.running = True
@@ -385,18 +382,14 @@ class ZenithPageserver(PgProtocol):
        self.initial_tenant = self.zenith_cli.run(['tenant', 'list']).stdout.strip()
        return self

-    def stop(self, immediate=False) -> 'ZenithPageserver':
+    def stop(self) -> 'ZenithPageserver':
        """
        Stop the page server.
        Returns self.
        """
-        cmd = ['stop']
-        if immediate:
-            cmd.append('immediate')

-        log.info(f"Stopping pageserver with {cmd}")
        if self.running:
-            self.zenith_cli.run(cmd)
+            self.zenith_cli.run(['stop'])
            self.running = False

        return self
@@ -405,7 +398,7 @@ class ZenithPageserver(PgProtocol):
        return self

    def __exit__(self, exc_type, exc, tb):
-        self.stop(True)
+        self.stop()

    @cached_property
    def auth_keys(self) -> AuthKeys:
@@ -426,7 +419,7 @@ class ZenithPageserver(PgProtocol):
 def pageserver_port(port_distributor: PortDistributor) -> PageserverPort:
    pg = port_distributor.get_port()
    http = port_distributor.get_port()
-    log.info(f"pageserver_port: pg={pg} http={http}")
+    print(f"pageserver_port: pg={pg} http={http}")
    return PageserverPort(pg=pg, http=http)


@@ -450,8 +443,8 @@ def pageserver(zenith_cli: ZenithCli, repo_dir: str, pageserver_port: Pageserver
    yield ps

    # After the yield comes any cleanup code we need.
-    log.info('Starting pageserver cleanup')
-    ps.stop(True)
+    print('Starting pageserver cleanup')
+    ps.stop()

 class PgBin:
    """ A helper class for executing postgres binaries """
@@ -488,7 +481,7 @@ class PgBin:
        """

        self._fixpath(command)
-        log.info('Running command "{}"'.format(' '.join(command)))
+        print('Running command "{}"'.format(' '.join(command)))
        env = self._build_env(env)
        subprocess.run(command, env=env, cwd=cwd, check=True)

@@ -505,7 +498,7 @@ class PgBin:
        """

        self._fixpath(command)
-        log.info('Running command "{}"'.format(' '.join(command)))
+        print('Running command "{}"'.format(' '.join(command)))
        env = self._build_env(env)
        return subprocess_capture(self.log_dir, command, env=env, cwd=cwd, check=True, **kwargs)

@@ -525,7 +518,7 @@ def pageserver_auth_enabled(zenith_cli: ZenithCli, repo_dir: str, pageserver_por
 class Postgres(PgProtocol):
    """ An object representing a running postgres daemon. """
    def __init__(self, zenith_cli: ZenithCli, repo_dir: str, pg_bin: PgBin, tenant_id: str, port: int):
-        super().__init__(host='127.0.0.1', port=port)
+        super().__init__(host='localhost', port=port)

        self.zenith_cli = zenith_cli
        self.running = False
@@ -573,12 +566,12 @@ class Postgres(PgProtocol):

        assert self.branch is not None

-        log.info(f"Starting postgres on branch {self.branch}")
+        print(f"Starting postgres on branch {self.branch}")

        run_result = self.zenith_cli.run(['pg', 'start', self.branch, f'--tenantid={self.tenant_id}', f'--port={self.port}'])
        self.running = True

-        log.info(f"stdout: {run_result.stdout}")
+        print(f"stdout: {run_result.stdout}")

        return self

@@ -793,7 +786,7 @@ def postgres(zenith_cli: ZenithCli, initial_tenant: str, repo_dir: str, pg_bin:
    yield pgfactory

    # After the yield comes any cleanup code we need.
-    log.info('Starting postgres cleanup')
+    print('Starting postgres cleanup')
    pgfactory.stop_all()

 def read_pid(path: Path):
@@ -823,36 +816,29 @@ class WalAcceptor:

        cmd = [str(self.wa_bin_path)]
        cmd.extend(["-D", str(self.data_dir)])
-        cmd.extend(["--listen-pg", f"127.0.0.1:{self.port.pg}"])
-        cmd.extend(["--listen-http", f"127.0.0.1:{self.port.http}"])
+        cmd.extend(["--listen-pg", f"localhost:{self.port.pg}"])
+        cmd.extend(["--listen-http", f"localhost:{self.port.http}"])
        cmd.append("--daemonize")
        cmd.append("--no-sync")
        # Tell page server it can receive WAL from this WAL safekeeper
-        cmd.extend(["--pageserver", f"127.0.0.1:{self.pageserver_port}"])
+        cmd.extend(["--pageserver", f"localhost:{self.pageserver_port}"])
        cmd.extend(["--recall", "1 second"])
-        log.info('Running command "{}"'.format(' '.join(cmd)))
+        print('Running command "{}"'.format(' '.join(cmd)))
        env = {'PAGESERVER_AUTH_TOKEN': self.auth_token} if self.auth_token else None
        subprocess.run(cmd, check=True, env=env)

-        # wait for wal acceptor start by checking its status
-        started_at = time.time()
-        while True:
-            try:
-                http_cli = self.http_client()
-                http_cli.check_status()
-            except Exception as e:
-                elapsed = time.time() - started_at
-                if elapsed > 3:
-                    raise RuntimeError(
-                        f"timed out waiting {elapsed:.0f}s for wal acceptor start: {e}")
-                time.sleep(0.5)
-            else:
-                break  # success
-        return self
+        # wait for wal acceptor start by checkking that pid is readable
+        for _ in range(3):
+            pid = self.get_pid()
+            if pid is not None:
+                return self
+            time.sleep(0.5)
+
+        raise RuntimeError("cannot get wal acceptor pid")

    @property
    def pidfile(self) -> Path:
-        return self.data_dir / "safekeeper.pid"
+        return self.data_dir / "wal_acceptor.pid"

    def get_pid(self) -> Optional[int]:
        if not self.pidfile.exists():
@@ -866,10 +852,10 @@ class WalAcceptor:
        return pid

    def stop(self) -> 'WalAcceptor':
-        log.info('Stopping wal acceptor {}'.format(self.num))
+        print('Stopping wal acceptor {}'.format(self.num))
        pid = self.get_pid()
        if pid is None:
-            log.info("Wal acceptor {} is not running".format(self.num))
+            print("Wal acceptor {} is not running".format(self.num))
            return self

        try:
@@ -888,27 +874,23 @@ class WalAcceptor:

        # "replication=0" hacks psycopg not to send additional queries
        # on startup, see https://github.com/psycopg/psycopg2/pull/482
-        connstr = f"host=127.0.0.1 port={self.port.pg} replication=0 options='-c ztimelineid={timeline_id} ztenantid={tenant_id}'"
+        connstr = f"host=localhost port={self.port.pg} replication=0 options='-c ztimelineid={timeline_id} ztenantid={tenant_id}'"

        with closing(psycopg2.connect(connstr)) as conn:
            # server doesn't support transactions
            conn.autocommit = True
            with conn.cursor() as cur:
                request_json = json.dumps(request)
-                log.info(f"JSON_CTRL request on port {self.port.pg}: {request_json}")
+                print(f"JSON_CTRL request on port {self.port.pg}: {request_json}")
                cur.execute("JSON_CTRL " + request_json)
                all = cur.fetchall()
-                log.info(f"JSON_CTRL response: {all[0][0]}")
+                print(f"JSON_CTRL response: {all[0][0]}")
                return json.loads(all[0][0])

-    def http_client(self):
-        return WalAcceptorHttpClient(port=self.port.http)
-
-
 class WalAcceptorFactory:
    """ An object representing multiple running wal acceptors. """
    def __init__(self, zenith_binpath: Path, data_dir: Path, pageserver_port: int, port_distributor: PortDistributor):
-        self.wa_bin_path = zenith_binpath / 'safekeeper'
+        self.wa_bin_path = zenith_binpath / 'wal_acceptor'
        self.data_dir = data_dir
        self.instances: List[WalAcceptor] = []
        self.port_distributor = port_distributor
@@ -949,7 +931,7 @@ class WalAcceptorFactory:

    def get_connstrs(self) -> str:
        """ Get list of wal acceptor endpoints suitable for wal_acceptors GUC  """
-        return ','.join(["127.0.0.1:{}".format(wa.port.pg) for wa in self.instances])
+        return ','.join(["localhost:{}".format(wa.port.pg) for wa in self.instances])


@zenfixture
@@ -963,34 +945,16 @@ def wa_factory(zenith_binpath: str, repo_dir: str, pageserver_port: PageserverPo
    )
    yield wafactory
    # After the yield comes any cleanup code we need.
-    log.info('Starting wal acceptors cleanup')
+    print('Starting wal acceptors cleanup')
    wafactory.stop_all()

-@dataclass
-class PageserverTimelineStatus:
-    acceptor_epoch: int
-
-class WalAcceptorHttpClient(requests.Session):
-    def __init__(self, port: int) -> None:
-        super().__init__()
-        self.port = port
-
-    def check_status(self):
-        self.get(f"http://127.0.0.1:{self.port}/v1/status").raise_for_status()
-
-    def timeline_status(self, tenant_id: str, timeline_id: str) -> PageserverTimelineStatus:
-        res = self.get(f"http://127.0.0.1:{self.port}/v1/timeline/{tenant_id}/{timeline_id}")
-        res.raise_for_status()
-        resj = res.json()
-        return PageserverTimelineStatus(acceptor_epoch=resj['acceptor_state']['epoch'])
-

@zenfixture
 def base_dir() -> str:
    """ find the base directory (currently this is the git root) """

    base_dir = os.path.normpath(os.path.join(get_self_dir(), '../..'))
-    log.info(f'base_dir is {base_dir}')
+    print('\nbase_dir is', base_dir)
    return base_dir


@@ -1019,7 +983,7 @@ def test_output_dir(request: Any, top_output_dir: str) -> str:
        test_name = 'shared'

    test_output_dir = os.path.join(top_output_dir, test_name)
-    log.info(f'test_output_dir is {test_output_dir}')
+    print('test_output_dir is', test_output_dir)
    shutil.rmtree(test_output_dir, ignore_errors=True)
    mkdir_if_needed(test_output_dir)
    return test_output_dir
@@ -1061,7 +1025,7 @@ def pg_distrib_dir(base_dir: str) -> str:
        pg_dir = env_postgres_bin
    else:
        pg_dir = os.path.normpath(os.path.join(base_dir, DEFAULT_POSTGRES_DIR))
-    log.info(f'postgres dir is {pg_dir}')
+    print('postgres dir is', pg_dir)
    if not os.path.exists(os.path.join(pg_dir, 'bin/postgres')):
        raise Exception('postgres not found at "{}"'.format(pg_dir))
    return pg_dir
@@ -1100,7 +1064,7 @@ def list_files_to_compare(pgdata_dir: str):
                pgdata_files.append(rel_file)

    pgdata_files.sort()
-    log.info(pgdata_files)
+    print(pgdata_files)
    return pgdata_files

 # pg is the existing and running compute node, that we want to compare with a basebackup
@@ -1124,7 +1088,7 @@ def check_restored_datadir_content(zenith_cli: ZenithCli, test_output_dir: str,
    cmd = rf"""
        {psql_path}                                    \
            --no-psqlrc                                \
-            postgres://127.0.0.1:{pageserver_pg_port}  \
+            postgres://localhost:{pageserver_pg_port}  \
            -c 'basebackup {pg.tenant_id} {timeline}'  \
         | tar -x -C {restored_dir_path}
    """
@@ -1146,7 +1110,9 @@ def check_restored_datadir_content(zenith_cli: ZenithCli, test_output_dir: str,
                                                restored_dir_path,
                                                pgdata_files,
                                                shallow=False)
-    log.info(f'filecmp result mismatch and error lists:\n\t mismatch={mismatch}\n\t error={error}')
+    print('filecmp result mismatch and error lists:')
+    print(mismatch)
+    print(error)

    for f in mismatch:

--- a/test_runner/netstat-script.sh
+++ b/test_runner/netstat-script.sh
@@ -1,8 +0,0 @@
-#!/bin/bash
-
-while true; do
-  echo -n "==== CURRENT TIME:" >> /tmp/test_output/netstat.stdout
-  date +"%T.%N" >> /tmp/test_output/netstat.stdout
-  sudo netstat -vpnoa | grep tcp | sort >> /tmp/test_output/netstat.stdout
-  sleep 0.5
-done
--- a/test_runner/performance/test_bulk_insert.py
+++ b/test_runner/performance/test_bulk_insert.py
@@ -1,7 +1,6 @@
 import os
 from contextlib import closing
 from fixtures.zenith_fixtures import PostgresFactory, ZenithPageserver
-from fixtures.log_helper import log

 pytest_plugins = ("fixtures.zenith_fixtures", "fixtures.benchmark_fixture")

@@ -20,7 +19,7 @@ def test_bulk_insert(postgres: PostgresFactory, pageserver: ZenithPageserver, pg
    zenith_cli.run(["branch", "test_bulk_insert", "empty"])

    pg = postgres.create_start('test_bulk_insert')
-    log.info("postgres is running on 'test_bulk_insert' branch")
+    print("postgres is running on 'test_bulk_insert' branch")

    # Open a connection directly to the page server that we'll use to force
    # flushing the layers to disk
--- a/test_runner/performance/test_gist_build.py
+++ b/test_runner/performance/test_gist_build.py
@@ -1,49 +0,0 @@
-import os
-from contextlib import closing
-from fixtures.zenith_fixtures import PostgresFactory, ZenithPageserver
-from fixtures.log_helper import log
-
-pytest_plugins = ("fixtures.zenith_fixtures", "fixtures.benchmark_fixture")
-
-#
-# Test buffering GisT build. It WAL-logs the whole relation, in 32-page chunks.
-# As of this writing, we're duplicate those giant WAL records for each page,
-# which makes the delta layer about 32x larger than it needs to be.
-#
-def test_gist_buffering_build(postgres: PostgresFactory, pageserver: ZenithPageserver, pg_bin, zenith_cli, zenbenchmark, repo_dir: str):
-    # Create a branch for us
-    zenith_cli.run(["branch", "test_gist_buffering_build", "empty"])
-
-    pg = postgres.create_start('test_gist_buffering_build')
-    log.info("postgres is running on 'test_gist_buffering_build' branch")
-
-    # Open a connection directly to the page server that we'll use to force
-    # flushing the layers to disk
-    psconn = pageserver.connect();
-    pscur = psconn.cursor()
-
-    # Get the timeline ID of our branch. We need it for the 'do_gc' command
-    with closing(pg.connect()) as conn:
-        with conn.cursor() as cur:
-            cur.execute("SHOW zenith.zenith_timeline")
-            timeline = cur.fetchone()[0]
-
-            # Create test table.
-            cur.execute("create table gist_point_tbl(id int4, p point)");
-            cur.execute("insert into gist_point_tbl select g, point(g, g) from generate_series(1, 1000000) g;");
-
-            # Build the index.
-            with zenbenchmark.record_pageserver_writes(pageserver, 'pageserver_writes'):
-                with zenbenchmark.record_duration('build'):
-                    cur.execute("create index gist_pointidx2 on gist_point_tbl using gist(p) with (buffering = on)");
-
-                    # Flush the layers from memory to disk. This is included in the reported
-                    # time and I/O
-                    pscur.execute(f"do_gc {pageserver.initial_tenant} {timeline} 1000000")
-
-            # Record peak memory usage
-            zenbenchmark.record("peak_mem", zenbenchmark.get_peak_mem(pageserver) / 1024, 'MB')
-
-            # Report disk space used by the repository
-            timeline_size = zenbenchmark.get_timeline_size(repo_dir, pageserver.initial_tenant, timeline)
-            zenbenchmark.record('size', timeline_size / (1024*1024), 'MB')
--- a/test_runner/performance/test_perf_pgbench.py
+++ b/test_runner/performance/test_perf_pgbench.py
@@ -1,7 +1,6 @@
 import os
 from contextlib import closing
 from fixtures.zenith_fixtures import PostgresFactory, ZenithPageserver
-from fixtures.log_helper import log

 pytest_plugins = ("fixtures.zenith_fixtures", "fixtures.benchmark_fixture")

@@ -19,7 +18,7 @@ def test_pgbench(postgres: PostgresFactory, pageserver: ZenithPageserver, pg_bin
    zenith_cli.run(["branch", "test_pgbench_perf", "empty"])

    pg = postgres.create_start('test_pgbench_perf')
-    log.info("postgres is running on 'test_pgbench_perf' branch")
+    print("postgres is running on 'test_pgbench_perf' branch")

    # Open a connection directly to the page server that we'll use to force
    # flushing the layers to disk
--- a/test_runner/performance/test_write_amplification.py
+++ b/test_runner/performance/test_write_amplification.py
@@ -13,7 +13,6 @@
 import os
 from contextlib import closing
 from fixtures.zenith_fixtures import PostgresFactory, ZenithPageserver
-from fixtures.log_helper import log

 pytest_plugins = ("fixtures.zenith_fixtures", "fixtures.benchmark_fixture")

@@ -22,7 +21,7 @@ def test_write_amplification(postgres: PostgresFactory, pageserver: ZenithPagese
    zenith_cli.run(["branch", "test_write_amplification", "empty"])

    pg = postgres.create_start('test_write_amplification')
-    log.info("postgres is running on 'test_write_amplification' branch")
+    print("postgres is running on 'test_write_amplification' branch")

    # Open a connection directly to the page server that we'll use to force
    # flushing the layers to disk
--- a/test_runner/pytest.ini
+++ b/test_runner/pytest.ini
@@ -1,4 +1,2 @@
 [pytest]
 minversion = 6.0
-log_format = %(asctime)s.%(msecs)-3d %(levelname)s [%(filename)s:%(lineno)d] %(message)s
-log_date_format = %Y-%m-%d %H:%M:%S
--- a/test_runner/test_broken.py
+++ b/test_runner/test_broken.py
@@ -1,8 +1,6 @@
 import pytest
 import os

-from fixtures.log_helper import log
-
 pytest_plugins = ("fixtures.zenith_fixtures")
 """
 Use this test to see what happens when tests fail.
@@ -24,7 +22,7 @@ def test_broken(zenith_cli, pageserver, postgres, pg_bin):
    zenith_cli.run(["branch", "test_broken", "empty"])

    postgres.create_start("test_broken")
-    log.info('postgres is running')
+    print('postgres is running')

-    log.info('THIS NEXT COMMAND WILL FAIL:')
+    print('THIS NEXT COMMAND WILL FAIL:')
    pg_bin.run('pgbench -i_am_a_broken_test'.split())
--- a/vendor/postgres
+++ b/vendor/postgres
--- a/walkeeper/Cargo.toml
+++ b/walkeeper/Cargo.toml
@@ -11,8 +11,6 @@ regex = "1.4.5"
 bincode = "1.3"
 bytes = "1.0.1"
 byteorder = "1.4.3"
-hyper = "0.14"
-routerify = "2"
 fs2 = "0.4.3"
 lazy_static = "1.4.0"
 serde_json = "1"
--- a/walkeeper/README
+++ b/walkeeper/README
@@ -89,12 +89,12 @@ A: Page Server is a single server which can be lost. As our primary

 Q: What if the compute node evicts a page, needs it back, but the page is yet
   to reach the Page Server?
-A: If the compute node has evicted a page, changes to it have been WAL-logged
-   (that's why it is called Write Ahead logging; there are some exceptions like
-   index builds, but these are exceptions). These WAL records will eventually
-   reach the Page Server. The Page Server notes that the compute note requests
-   pages with a very recent LSN and will not respond to the compute node until a
-   corresponding WAL is received from WAL safekeepers.
+A: If the compute node has evicted a page, all changes from that page are
+   already committed, i.e. they are saved on majority of WAL safekeepers. These
+   WAL records will eventually reach the Page Server. The Page Server notes
+   that the compute note requests pages with a very recent LSN and will not
+   respond to the compute node until it a corresponding WAL is received from WAL
+   safekeepers.

 Q: How long may Page Server wait for?
 A: Not too long, hopefully. If a page is evicted, it probably was not used for
--- a/walkeeper/src/bin/wal_acceptor.rs
+++ b/walkeeper/src/bin/wal_acceptor.rs
@@ -1,5 +1,5 @@
 //
-// Main entry point for the safekeeper executable
+// Main entry point for the wal_acceptor executable
 //
 use anyhow::Result;
 use clap::{App, Arg};
@@ -14,21 +14,20 @@ use zenith_utils::http::endpoint;
 use zenith_utils::logging;

 use walkeeper::defaults::{DEFAULT_HTTP_LISTEN_ADDR, DEFAULT_PG_LISTEN_ADDR};
-use walkeeper::http;
 use walkeeper::s3_offload;
 use walkeeper::wal_service;
 use walkeeper::WalAcceptorConf;

 fn main() -> Result<()> {
    zenith_metrics::set_common_metrics_prefix("safekeeper");
-    let arg_matches = App::new("Zenith safekeeper")
+    let arg_matches = App::new("Zenith wal_acceptor")
        .about("Store WAL stream to local file system and push it to WAL receivers")
        .arg(
            Arg::with_name("datadir")
                .short("D")
                .long("dir")
                .takes_value(true)
-                .help("Path to the safekeeper data directory"),
+                .help("Path to the WAL acceptor data directory"),
        )
        .arg(
            Arg::with_name("listen-pg")
@@ -129,20 +128,14 @@ fn main() -> Result<()> {
 }

 fn start_wal_acceptor(conf: WalAcceptorConf) -> Result<()> {
-    let log_filename = conf.data_dir.join("safekeeper.log");
-    let log_file = logging::init(log_filename, conf.daemonize)?;
+    let log_filename = conf.data_dir.join("wal_acceptor.log");
+    let (_scope_guard, log_file) = logging::init(log_filename, conf.daemonize)?;

    let http_listener = TcpListener::bind(conf.listen_http_addr.clone()).map_err(|e| {
        error!("failed to bind to address {}: {}", conf.listen_http_addr, e);
        e
    })?;

-    info!("Starting safekeeper on {}", conf.listen_pg_addr);
-    let pg_listener = TcpListener::bind(conf.listen_pg_addr.clone()).map_err(|e| {
-        error!("failed to bind to address {}: {}", conf.listen_pg_addr, e);
-        e
-    })?;
-
    if conf.daemonize {
        info!("daemonizing...");

@@ -152,7 +145,7 @@ fn start_wal_acceptor(conf: WalAcceptorConf) -> Result<()> {
        let stderr = log_file;

        let daemonize = Daemonize::new()
-            .pid_file("safekeeper.pid")
+            .pid_file("wal_acceptor.pid")
            .working_directory(Path::new("."))
            .stdout(stdout)
            .stderr(stderr);
@@ -165,12 +158,11 @@ fn start_wal_acceptor(conf: WalAcceptorConf) -> Result<()> {

    let mut threads = Vec::new();

-    let conf_cloned = conf.clone();
    let http_endpoint_thread = thread::Builder::new()
        .name("http_endpoint_thread".into())
-        .spawn(|| {
-            // TODO authentication
-            let router = http::make_router(conf_cloned);
+        .spawn(move || {
+            // No authentication at all: read-only metrics only, early stage.
+            let router = endpoint::make_router();
            endpoint::serve_thread_main(router, http_listener).unwrap();
        })
        .unwrap();
@@ -192,7 +184,7 @@ fn start_wal_acceptor(conf: WalAcceptorConf) -> Result<()> {
        .name("WAL acceptor thread".into())
        .spawn(|| {
            // thread code
-            let thread_result = wal_service::thread_main(conf, pg_listener);
+            let thread_result = wal_service::thread_main(conf);
            if let Err(e) = thread_result {
                info!("wal_service thread terminated: {}", e);
            }
--- a/walkeeper/src/http/mod.rs
+++ b/walkeeper/src/http/mod.rs
@@ -1,2 +0,0 @@
-pub mod routes;
-pub use routes::make_router;
--- a/walkeeper/src/http/routes.rs
+++ b/walkeeper/src/http/routes.rs
@@ -1,88 +0,0 @@
-use hyper::{Body, Request, Response, StatusCode};
-use routerify::ext::RequestExt;
-use routerify::RouterBuilder;
-use serde::Serialize;
-use serde::Serializer;
-use std::fmt::Display;
-use std::sync::Arc;
-use zenith_utils::lsn::Lsn;
-
-use crate::safekeeper::AcceptorState;
-use crate::timeline::CreateControlFile;
-use crate::timeline::GlobalTimelines;
-use crate::WalAcceptorConf;
-use zenith_utils::http::endpoint;
-use zenith_utils::http::error::ApiError;
-use zenith_utils::http::json::json_response;
-use zenith_utils::http::request::parse_request_param;
-use zenith_utils::zid::{ZTenantId, ZTimelineId};
-
-/// Healthcheck handler.
-async fn status_handler(_: Request<Body>) -> Result<Response<Body>, ApiError> {
-    Ok(json_response(StatusCode::OK, "")?)
-}
-
-fn get_conf(request: &Request<Body>) -> &WalAcceptorConf {
-    request
-        .data::<Arc<WalAcceptorConf>>()
-        .expect("unknown state type")
-        .as_ref()
-}
-
-fn display_serialize<S, F>(z: &F, s: S) -> Result<S::Ok, S::Error>
-where
-    S: Serializer,
-    F: Display,
-{
-    s.serialize_str(&format!("{}", z))
-}
-
-/// Info about timeline on safekeeper ready for reporting.
-#[derive(Debug, Serialize)]
-struct TimelineStatus {
-    #[serde(serialize_with = "display_serialize")]
-    tenant_id: ZTenantId,
-    #[serde(serialize_with = "display_serialize")]
-    timeline_id: ZTimelineId,
-    acceptor_state: AcceptorState,
-    #[serde(serialize_with = "display_serialize")]
-    commit_lsn: Lsn,
-    #[serde(serialize_with = "display_serialize")]
-    truncate_lsn: Lsn,
-}
-
-/// Report info about timeline.
-async fn timeline_status_handler(request: Request<Body>) -> Result<Response<Body>, ApiError> {
-    let tenant_id: ZTenantId = parse_request_param(&request, "tenant_id")?;
-    let timeline_id: ZTimelineId = parse_request_param(&request, "timeline_id")?;
-
-    let tli = GlobalTimelines::get(
-        get_conf(&request),
-        tenant_id,
-        timeline_id,
-        CreateControlFile::False,
-    )
-    .map_err(ApiError::from_err)?;
-    let sk_state = tli.get_info();
-
-    let status = TimelineStatus {
-        tenant_id,
-        timeline_id,
-        acceptor_state: sk_state.acceptor_state,
-        commit_lsn: sk_state.commit_lsn,
-        truncate_lsn: sk_state.truncate_lsn,
-    };
-    Ok(json_response(StatusCode::OK, status)?)
-}
-
-/// Safekeeper http router.
-pub fn make_router(conf: WalAcceptorConf) -> RouterBuilder<hyper::Body, ApiError> {
-    let router = endpoint::make_router();
-    router
-        .data(Arc::new(conf))
-        .get("/v1/status", status_handler)
-        .get(
-            "/v1/timeline/:tenant_id/:timeline_id",
-            timeline_status_handler,
-        )
-}
--- a/walkeeper/src/lib.rs
+++ b/walkeeper/src/lib.rs
@@ -2,7 +2,6 @@
 use std::path::PathBuf;
 use std::time::Duration;

-pub mod http;
 pub mod json_ctrl;
 pub mod receive_wal;
 pub mod replication;
--- a/walkeeper/src/safekeeper.rs
+++ b/walkeeper/src/safekeeper.rs
@@ -290,13 +290,13 @@ lazy_static! {
        "Current flush_lsn, grouped by timeline",
        &["ztli"]
    )
-    .expect("Failed to register safekeeper_flush_lsn gauge vec");
+    .expect("Failed to register safekeeper_flush_lsn int gauge vec");
    static ref COMMIT_LSN_GAUGE: GaugeVec = register_gauge_vec!(
        "safekeeper_commit_lsn",
        "Current commit_lsn (not necessarily persisted to disk), grouped by timeline",
        &["ztli"]
    )
-    .expect("Failed to register safekeeper_commit_lsn gauge vec");
+    .expect("Failed to register safekeeper_commit_lsn int gauge vec");
 }

 struct SafeKeeperMetrics {
@@ -312,13 +312,6 @@ impl SafeKeeperMetrics {
            commit_lsn: COMMIT_LSN_GAUGE.with_label_values(&[&ztli_str]),
        }
    }
-
-    fn new_noname() -> SafeKeeperMetrics {
-        SafeKeeperMetrics {
-            flush_lsn: FLUSH_LSN_GAUGE.with_label_values(&["n/a"]),
-            commit_lsn: COMMIT_LSN_GAUGE.with_label_values(&["n/a"]),
-        }
-    }
 }

 /// SafeKeeper which consumes events (messages from compute) and provides
@@ -329,7 +322,7 @@ pub struct SafeKeeper<ST: Storage> {
    pub flush_lsn: Lsn,
    pub tli: u32,
    // Cached metrics so we don't have to recompute labels on each update.
-    metrics: SafeKeeperMetrics,
+    metrics: Option<SafeKeeperMetrics>,
    /// not-yet-flushed pairs of same named fields in s.*
    pub commit_lsn: Lsn,
    pub truncate_lsn: Lsn,
@@ -348,7 +341,7 @@ where
        SafeKeeper {
            flush_lsn,
            tli,
-            metrics: SafeKeeperMetrics::new_noname(),
+            metrics: None,
            commit_lsn: state.commit_lsn,
            truncate_lsn: state.truncate_lsn,
            storage,
@@ -400,7 +393,7 @@ where
        self.s.server.wal_seg_size = msg.wal_seg_size;
        self.storage.persist(&self.s, true)?;

-        self.metrics = SafeKeeperMetrics::new(self.s.server.ztli);
+        self.metrics = Some(SafeKeeperMetrics::new(self.s.server.ztli));

        info!(
            "processed greeting from proposer {:?}, sending term {:?}",
@@ -525,7 +518,11 @@ where
        }
        if last_rec_lsn > self.flush_lsn {
            self.flush_lsn = last_rec_lsn;
-            self.metrics.flush_lsn.set(u64::from(self.flush_lsn) as f64);
+            self.metrics
+                .as_ref()
+                .unwrap()
+                .flush_lsn
+                .set(u64::from(self.flush_lsn) as f64);
        }

        // Advance commit_lsn taking into account what we have locally. xxx this
@@ -544,6 +541,8 @@ where
                commit_lsn >= msg.h.epoch_start_lsn && self.s.commit_lsn < msg.h.epoch_start_lsn;
            self.commit_lsn = commit_lsn;
            self.metrics
+                .as_ref()
+                .unwrap()
                .commit_lsn
                .set(u64::from(self.commit_lsn) as f64);
        }
--- a/walkeeper/src/send_wal.rs
+++ b/walkeeper/src/send_wal.rs
@@ -13,13 +13,14 @@ use std::str::FromStr;
 use std::sync::Arc;
 use zenith_utils::postgres_backend;
 use zenith_utils::postgres_backend::PostgresBackend;
-use zenith_utils::pq_proto::{BeMessage, FeStartupMessage, RowDescriptor, INT4_OID, TEXT_OID};
+use zenith_utils::pq_proto::{BeMessage, FeStartupMessage, RowDescriptor};
 use zenith_utils::zid::{ZTenantId, ZTimelineId};

 use crate::timeline::CreateControlFile;

 /// Handler for streaming WAL from acceptor
 pub struct SendWalHandler {
+    /// wal acceptor configuration
    pub conf: WalAcceptorConf,
    /// assigned application name
    pub appname: Option<String>,
@@ -71,16 +72,19 @@ impl postgres_backend::Handler for SendWalHandler {
        }
        if query_string.starts_with(b"IDENTIFY_SYSTEM") {
            self.handle_identify_system(pgb)?;
+            Ok(())
        } else if query_string.starts_with(b"START_REPLICATION") {
            ReplicationConn::new(pgb).run(self, pgb, &query_string)?;
+            Ok(())
        } else if query_string.starts_with(b"START_WAL_PUSH") {
            ReceiveWalConn::new(pgb)?.run(self)?;
+            Ok(())
        } else if query_string.starts_with(b"JSON_CTRL") {
            handle_json_ctrl(self, pgb, &query_string)?;
+            Ok(())
        } else {
            bail!("Unexpected command {:?}", query_string);
        }
-        Ok(())
    }
 }

@@ -110,25 +114,25 @@ impl SendWalHandler {
        pgb.write_message_noflush(&BeMessage::RowDescription(&[
            RowDescriptor {
                name: b"systemid",
-                typoid: TEXT_OID,
+                typoid: 25,
                typlen: -1,
                ..Default::default()
            },
            RowDescriptor {
                name: b"timeline",
-                typoid: INT4_OID,
+                typoid: 23,
                typlen: 4,
                ..Default::default()
            },
            RowDescriptor {
                name: b"xlogpos",
-                typoid: TEXT_OID,
+                typoid: 25,
                typlen: -1,
                ..Default::default()
            },
            RowDescriptor {
                name: b"dbname",
-                typoid: TEXT_OID,
+                typoid: 25,
                typlen: -1,
                ..Default::default()
            },
--- a/walkeeper/src/timeline.rs
+++ b/walkeeper/src/timeline.rs
@@ -112,7 +112,7 @@ impl SharedState {
        }
        match opts.open(&control_file_path) {
            Ok(mut file) => {
-                // Lock file to prevent two or more active safekeepers
+                // Lock file to prevent two or more active wal_acceptors
                match file.try_lock_exclusive() {
                    Ok(()) => {}
                    Err(e) => {
@@ -289,7 +289,7 @@ lazy_static! {
 }

 /// A zero-sized struct used to manage access to the global timelines map.
-pub struct GlobalTimelines;
+struct GlobalTimelines;

 impl GlobalTimelines {
    /// Get a timeline with control file loaded from the global TIMELINES map.
--- a/walkeeper/src/wal_service.rs
+++ b/walkeeper/src/wal_service.rs
@@ -12,7 +12,13 @@ use crate::WalAcceptorConf;
 use zenith_utils::postgres_backend::{AuthType, PostgresBackend};

 /// Accept incoming TCP connections and spawn them into a background thread.
-pub fn thread_main(conf: WalAcceptorConf, listener: TcpListener) -> Result<()> {
+pub fn thread_main(conf: WalAcceptorConf) -> Result<()> {
+    info!("Starting wal acceptor on {}", conf.listen_pg_addr);
+    let listener = TcpListener::bind(conf.listen_pg_addr.clone()).map_err(|e| {
+        error!("failed to bind to address {}: {}", conf.listen_pg_addr, e);
+        e
+    })?;
+
    loop {
        match listener.accept() {
            Ok((socket, peer_addr)) => {
@@ -35,8 +41,8 @@ fn handle_socket(socket: TcpStream, conf: WalAcceptorConf) -> Result<()> {
    socket.set_nodelay(true)?;

    let mut conn_handler = SendWalHandler::new(conf);
-    let pgbackend = PostgresBackend::new(socket, AuthType::Trust, None, false)?;
-    // libpq replication protocol between safekeeper and replicas/pagers
+    let pgbackend = PostgresBackend::new(socket, AuthType::Trust, None)?;
+    // libpq replication protocol between wal_acceptor and replicas/pagers
    pgbackend.run(&mut conn_handler)?;

    Ok(())
--- a/zenith/src/main.rs
+++ b/zenith/src/main.rs
@@ -88,12 +88,7 @@ fn main() -> Result<()> {
        )
        .subcommand(SubCommand::with_name("status"))
        .subcommand(SubCommand::with_name("start").about("Start local pageserver"))
-        .subcommand(SubCommand::with_name("stop").about("Stop local pageserver")
-                    .arg(Arg::with_name("immediate")
-                    .help("Don't flush repository data at shutdown")
-                    .required(false)
-                    )
-        )
+        .subcommand(SubCommand::with_name("stop").about("Stop local pageserver"))
        .subcommand(SubCommand::with_name("restart").about("Restart local pageserver"))
        .subcommand(
            SubCommand::with_name("pg")
@@ -201,12 +196,10 @@ fn main() -> Result<()> {
            }
        }

-        ("stop", Some(stop_match)) => {
+        ("stop", Some(_sub_m)) => {
            let pageserver = PageServerNode::from_env(&env);

-            let immediate = stop_match.is_present("immediate");
-
-            if let Err(e) = pageserver.stop(immediate) {
+            if let Err(e) = pageserver.stop() {
                eprintln!("pageserver stop failed: {}", e);
                exit(1);
            }
@@ -215,8 +208,7 @@ fn main() -> Result<()> {
        ("restart", Some(_sub_m)) => {
            let pageserver = PageServerNode::from_env(&env);

-            //TODO what shutdown strategy should we use here?
-            if let Err(e) = pageserver.stop(false) {
+            if let Err(e) = pageserver.stop() {
                eprintln!("pageserver stop failed: {}", e);
                exit(1);
            }
--- a/zenith_utils/Cargo.toml
+++ b/zenith_utils/Cargo.toml
@@ -18,9 +18,13 @@ serde = { version = "1.0", features = ["derive"] }
 serde_json = "1"
 thiserror = "1.0"
 tokio = "1.11"
-tracing = "0.1"
-tracing-log = "0.1"
-tracing-subscriber = "0.2"
+crossbeam-utils = "0.8.5"
+
+slog-async = "2.6.0"
+slog-stdlog = "4.1.0"
+slog-scope = "4.4.0"
+slog-term = "2.8.0"
+slog = "2.7.0"

 zenith_metrics = { path = "../zenith_metrics" }
 workspace_hack = { path = "../workspace_hack" }
--- a/zenith_utils/src/batch_fsync.rs
+++ b/zenith_utils/src/batch_fsync.rs
@@ -0,0 +1,45 @@
+use crossbeam_utils::thread;
+
+use std::{
+    fs::File,
+    io,
+    path::PathBuf,
+    sync::atomic::{AtomicUsize, Ordering},
+};
+
+pub fn batch_fsync(paths: &[PathBuf]) -> std::io::Result<()> {
+    let next = AtomicUsize::new(0);
+
+    let num_threads = std::cmp::min(paths.len() / 2, 256);
+
+    if num_threads <= 1 {
+        for path in paths {
+            let file = File::open(&path)?;
+            file.sync_all()?;
+        }
+
+        return Ok(());
+    }
+
+    thread::scope(|s| -> io::Result<()> {
+        let mut handles = Vec::new();
+        for _ in 0..num_threads {
+            handles.push(s.spawn(|_| loop {
+                let idx = next.fetch_add(1, Ordering::Relaxed);
+                if idx >= paths.len() {
+                    return io::Result::Ok(());
+                }
+
+                let file = File::open(&paths[idx])?;
+                file.sync_all()?;
+            }));
+        }
+
+        for handle in handles {
+            handle.join().unwrap()?;
+        }
+
+        Ok(())
+    })
+    .unwrap()
+}
--- a/zenith_utils/src/connstring.rs
+++ b/zenith_utils/src/connstring.rs
@@ -29,24 +29,24 @@ mod tests {

    #[test]
    fn test_connection_host_port() {
-        let config: Config = "postgresql://no_user@127.0.0.1:64000/no_db"
+        let config: Config = "postgresql://no_user@localhost:64000/no_db"
            .parse()
            .unwrap();
        assert_eq!(
            connection_host_port(&config),
-            ("127.0.0.1".to_owned(), 64000)
+            ("localhost".to_owned(), 64000)
        );
    }

    #[test]
    #[should_panic(expected = "only one pair of host and port is supported in connection string")]
    fn test_connection_host_port_multiple_ports() {
-        let config: Config = "postgresql://no_user@127.0.0.1:64000,127.0.0.1:64001/no_db"
+        let config: Config = "postgresql://no_user@localhost:64000,localhost:64001/no_db"
            .parse()
            .unwrap();
        assert_eq!(
            connection_host_port(&config),
-            ("127.0.0.1".to_owned(), 64000)
+            ("localhost".to_owned(), 64000)
        );
    }
 }
--- a/zenith_utils/src/http/endpoint.rs
+++ b/zenith_utils/src/http/endpoint.rs
@@ -12,17 +12,8 @@ use std::net::TcpListener;
 use zenith_metrics::{new_common_metric_name, register_int_counter, IntCounter};
 use zenith_metrics::{Encoder, TextEncoder};

-use std::sync::Mutex;
-use tokio::sync::oneshot::Sender;
-
 use super::error::ApiError;

-lazy_static! {
-    /// Channel used to send shutdown signal - wrapped in an Option to allow
-    /// it to be taken by value (since oneshot channels consume themselves on send)
-    static ref SHUTDOWN_SENDER: Mutex<Option<Sender<()>>> = Mutex::new(None);
-}
-
 lazy_static! {
    static ref SERVE_METRICS_COUNT: IntCounter = register_int_counter!(
        new_common_metric_name("serve_metrics_count"),
@@ -152,18 +143,11 @@ pub fn check_permission(req: &Request<Body>, tenantid: Option<ZTenantId>) -> Res
    }
 }

-// Send shutdown signal
-pub fn shutdown() {
-    if let Some(tx) = SHUTDOWN_SENDER.lock().unwrap().take() {
-        let _ = tx.send(());
-    }
-}
-
 pub fn serve_thread_main(
    router_builder: RouterBuilder<hyper::Body, ApiError>,
    listener: TcpListener,
 ) -> anyhow::Result<()> {
-    log::info!("Starting a http endpoint at {}", listener.local_addr()?);
+    log::info!("Starting a http endoint at {}", listener.local_addr()?);

    // Create a Service from the router above to handle incoming requests.
    let service = RouterService::new(router_builder.build().map_err(|err| anyhow!(err))?).unwrap();
@@ -175,14 +159,7 @@ pub fn serve_thread_main(

    let _guard = runtime.enter();

-    let (send, recv) = tokio::sync::oneshot::channel::<()>();
-    *SHUTDOWN_SENDER.lock().unwrap() = Some(send);
-
-    let server = Server::from_tcp(listener)?
-        .serve(service)
-        .with_graceful_shutdown(async {
-            recv.await.ok();
-        });
+    let server = Server::from_tcp(listener)?.serve(service);

    runtime.block_on(server)?;

--- a/zenith_utils/src/http/mod.rs
+++ b/zenith_utils/src/http/mod.rs
@@ -1,4 +1,3 @@
 pub mod endpoint;
 pub mod error;
 pub mod json;
-pub mod request;
--- a/zenith_utils/src/http/request.rs
+++ b/zenith_utils/src/http/request.rs
@@ -1,33 +0,0 @@
-use std::str::FromStr;
-
-use super::error::ApiError;
-use hyper::{Body, Request};
-use routerify::ext::RequestExt;
-
-pub fn get_request_param<'a>(
-    request: &'a Request<Body>,
-    param_name: &str,
-) -> Result<&'a str, ApiError> {
-    match request.param(param_name) {
-        Some(arg) => Ok(arg),
-        None => {
-            return Err(ApiError::BadRequest(format!(
-                "no {} specified in path param",
-                param_name
-            )))
-        }
-    }
-}
-
-pub fn parse_request_param<T: FromStr>(
-    request: &Request<Body>,
-    param_name: &str,
-) -> Result<T, ApiError> {
-    match get_request_param(request, param_name)?.parse() {
-        Ok(v) => Ok(v),
-        Err(_) => Err(ApiError::BadRequest(format!(
-            "failed to parse {}",
-            param_name
-        ))),
-    }
-}
--- a/zenith_utils/src/lib.rs
+++ b/zenith_utils/src/lib.rs
@@ -40,3 +40,6 @@ pub mod logging;

 // Misc
 pub mod accum;
+
+/// Utility for quickly fsyncing many files at once
+pub mod batch_fsync;
--- a/zenith_utils/src/logging.rs
+++ b/zenith_utils/src/logging.rs
@@ -1,3 +1,4 @@
+use slog::{Drain, Level};
 use std::{
    fs::{File, OpenOptions},
    path::Path,
@@ -5,12 +6,10 @@ use std::{

 use anyhow::{Context, Result};

-use tracing::subscriber::set_global_default;
-use tracing_log::LogTracer;
-use tracing_subscriber::fmt;
-use tracing_subscriber::{layer::SubscriberExt, EnvFilter, Registry};
-
-pub fn init(log_filename: impl AsRef<Path>, daemonize: bool) -> Result<File> {
+pub fn init(
+    log_filename: impl AsRef<Path>,
+    daemonize: bool,
+) -> Result<(slog_scope::GlobalLoggerGuard, File)> {
    // Don't open the same file for output multiple times;
    // the different fds could overwrite each other's output.
    let log_file = OpenOptions::new()
@@ -19,38 +18,30 @@ pub fn init(log_filename: impl AsRef<Path>, daemonize: bool) -> Result<File> {
        .open(&log_filename)
        .with_context(|| format!("failed to open {:?}", log_filename.as_ref()))?;

-    let default_filter_str = "info";
-
-    // We fall back to printing all spans at info-level or above if
-    // the RUST_LOG environment variable is not set.
-    let env_filter =
-        EnvFilter::try_from_default_env().unwrap_or_else(|_| EnvFilter::new(default_filter_str));
-
    // we are cloning and returning log file in order to allow redirecting daemonized stdout and stderr to it
    // if we do not use daemonization (e.g. in docker) it is better to log to stdout directly
    // for example to be in line with docker log command which expects logs comimg from stdout
-    //
-    // TODO: perhaps use a more human-readable format when !daemonize
-    if daemonize {
-        let x = log_file.try_clone().unwrap();
-
-        let fmt_layer = fmt::layer()
-            .pretty()
-            .with_target(false) // don't include event targets
-            .with_ansi(false) // don't use colors in log file
-            .with_writer(move || x.try_clone().unwrap());
-        let subscriber = Registry::default().with(env_filter).with(fmt_layer);
-
-        set_global_default(subscriber).expect("Failed to set subscriber");
+    let guard = if daemonize {
+        let decorator = slog_term::PlainSyncDecorator::new(log_file.try_clone()?);
+        let drain = slog_term::FullFormat::new(decorator)
+            .build()
+            .filter_level(Level::Info)
+            .fuse();
+        let logger = slog::Logger::root(drain, slog::o!());
+        slog_scope::set_global_logger(logger)
    } else {
-        let fmt_layer = fmt::layer().with_target(false); // don't include event targets
-        let subscriber = Registry::default().with(env_filter).with(fmt_layer);
+        let decorator = slog_term::TermDecorator::new().build();
+        let drain = slog_term::FullFormat::new(decorator)
+            .build()
+            .filter_level(Level::Info)
+            .fuse();
+        let drain = slog_async::Async::new(drain).chan_size(1000).build().fuse();
+        let logger = slog::Logger::root(drain, slog::o!());
+        slog_scope::set_global_logger(logger)
+    };

-        set_global_default(subscriber).expect("Failed to set subscriber");
-    }
+    // initialise forwarding of std log calls
+    slog_stdlog::init()?;

-    // Redirect all `log`'s events to our subscriber
-    LogTracer::init().expect("Failed to set logger");
-
-    Ok(log_file)
+    Ok((guard, log_file))
 }
--- a/zenith_utils/src/postgres_backend.rs
+++ b/zenith_utils/src/postgres_backend.rs
@@ -13,11 +13,7 @@ use serde::{Deserialize, Serialize};
 use std::io::{self, Write};
 use std::net::{Shutdown, SocketAddr, TcpStream};
 use std::str::FromStr;
-use std::sync::atomic::{AtomicBool, Ordering};
 use std::sync::Arc;
-use std::time::Duration;
-
-static PGBACKEND_SHUTDOWN_REQUESTED: AtomicBool = AtomicBool::new(false);

 pub trait Handler {
    /// Handle single query.
@@ -139,32 +135,13 @@ pub fn query_from_cstring(query_string: Bytes) -> Vec<u8> {
    query_string
 }

-// Helper function for socket read loops
-pub fn is_socket_read_timed_out(error: &anyhow::Error) -> bool {
-    for cause in error.chain() {
-        if let Some(io_error) = cause.downcast_ref::<io::Error>() {
-            if io_error.kind() == std::io::ErrorKind::WouldBlock {
-                return true;
-            }
-        }
-    }
-    false
-}
-
 impl PostgresBackend {
    pub fn new(
        socket: TcpStream,
        auth_type: AuthType,
        tls_config: Option<Arc<rustls::ServerConfig>>,
-        set_read_timeout: bool,
    ) -> io::Result<Self> {
        let peer_addr = socket.peer_addr()?;
-        if set_read_timeout {
-            socket
-                .set_read_timeout(Some(Duration::from_secs(5)))
-                .unwrap();
-        }
-
        Ok(Self {
            stream: Some(Stream::Bidirectional(BidiStream::from_tcp(socket))),
            buf_out: BytesMut::with_capacity(10 * 1024),
@@ -252,26 +229,12 @@ impl PostgresBackend {

        let mut unnamed_query_string = Bytes::new();

-        while !PGBACKEND_SHUTDOWN_REQUESTED.load(Ordering::Relaxed) {
-            match self.read_message() {
-                Ok(message) => {
-                    if let Some(msg) = message {
-                        trace!("got message {:?}", msg);
+        while let Some(msg) = self.read_message()? {
+            trace!("got message {:?}", msg);

-                        match self.process_message(handler, msg, &mut unnamed_query_string)? {
-                            ProcessMsgResult::Continue => continue,
-                            ProcessMsgResult::Break => break,
-                        }
-                    } else {
-                        break;
-                    }
-                }
-                Err(e) => {
-                    // If it is a timeout error, continue the loop
-                    if !is_socket_read_timed_out(&e) {
-                        return Err(e);
-                    }
-                }
+            match self.process_message(handler, msg, &mut unnamed_query_string)? {
+                ProcessMsgResult::Continue => continue,
+                ProcessMsgResult::Break => break,
            }
        }

@@ -464,8 +427,3 @@ impl PostgresBackend {
        Ok(ProcessMsgResult::Continue)
    }
 }
-
-// Set the flag to inform connections to cancel
-pub fn set_pgbackend_shutdown_requested() {
-    PGBACKEND_SHUTDOWN_REQUESTED.swap(true, Ordering::Relaxed);
-}
--- a/zenith_utils/src/pq_proto.rs
+++ b/zenith_utils/src/pq_proto.rs
@@ -15,9 +15,8 @@ use std::str;
 pub type Oid = u32;
 pub type SystemId = u64;

-pub const INT8_OID: Oid = 20;
-pub const INT4_OID: Oid = 23;
 pub const TEXT_OID: Oid = 25;
+pub const INT8_OID: Oid = 20;

 #[derive(Debug)]
 pub enum FeMessage {
--- a/zenith_utils/tests/ssl_test.rs
+++ b/zenith_utils/tests/ssl_test.rs
@@ -110,7 +110,7 @@ fn ssl() {
        .unwrap();
    let tls_config = Some(Arc::new(cfg));

-    let pgb = PostgresBackend::new(server_sock, AuthType::Trust, tls_config, true).unwrap();
+    let pgb = PostgresBackend::new(server_sock, AuthType::Trust, tls_config).unwrap();
    pgb.run(&mut handler).unwrap();
    assert!(handler.got_query);

@@ -150,7 +150,7 @@ fn no_ssl() {

    let mut handler = TestHandler;

-    let pgb = PostgresBackend::new(server_sock, AuthType::Trust, None, true).unwrap();
+    let pgb = PostgresBackend::new(server_sock, AuthType::Trust, None).unwrap();
    pgb.run(&mut handler).unwrap();

    client_jh.join().unwrap();
@@ -214,7 +214,7 @@ fn server_forces_ssl() {
        .unwrap();
    let tls_config = Some(Arc::new(cfg));

-    let pgb = PostgresBackend::new(server_sock, AuthType::Trust, tls_config, true).unwrap();
+    let pgb = PostgresBackend::new(server_sock, AuthType::Trust, tls_config).unwrap();
    let res = pgb.run(&mut handler).unwrap_err();
    assert_eq!("client did not connect with TLS", format!("{}", res));