Fix bug in using brin index in GC

Became actual after #674

.circleci/config.yml

@@ -7,7 +7,7 @@ executors:
   zenith-build-executor:
     resource_class: xlarge
     docker:
-      - image: cimg/rust:1.52.1
+      - image: cimg/rust:1.55.0
 
 jobs:
   check-codestyle:
@@ -110,7 +110,7 @@ jobs:
             # Require an exact match. While an out of date cache might speed up the build,
             # there's no way to clean out old packages, so the cache grows every time something
             # changes.
-            - v03-rust-cache-deps-<< parameters.build_type >>-{{ checksum "Cargo.lock" }}
+            - v04-rust-cache-deps-<< parameters.build_type >>-{{ checksum "Cargo.lock" }}
 
         # Build the rust code, including test binaries
       - run:
@@ -128,7 +128,7 @@ jobs:
 
       - save_cache:
           name: Save rust cache
-          key: v03-rust-cache-deps-<< parameters.build_type >>-{{ checksum "Cargo.lock" }}
+          key: v04-rust-cache-deps-<< parameters.build_type >>-{{ checksum "Cargo.lock" }}
           paths:
             - ~/.cargo/registry
             - ~/.cargo/git
@@ -182,6 +182,21 @@ jobs:
           paths:
             - "*"
 
+  check-python:
+    executor: python/default
+    steps:
+      - checkout
+      - run:
+          name: Install pipenv & deps
+          working_directory: test_runner
+          command: |
+            pip install pipenv
+            pipenv install --dev
+      - run:
+          name: Run yapf to ensure code format
+          working_directory: test_runner
+          command: pipenv run yapf --recursive --diff .
+
   run-pytest:
     #description: "Run pytest"
     executor: python/default
@@ -241,8 +256,6 @@ 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
@@ -254,8 +267,6 @@ jobs:
             # -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
       - 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
@@ -337,6 +348,7 @@ workflows:
   build_and_test:
     jobs:
       - check-codestyle
+      - check-python
       - build-postgres:
           name: build-postgres-<< matrix.build_type >>
           matrix:

Cargo.lock

@@ -918,6 +918,15 @@ dependencies = [
  "cfg-if 1.0.0",
 ]
 
+[[package]]
+name = "lz4_flex"
+version = "0.9.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "177c079243f6867429aca5af5053747f57e329d44f0c58bebca078cd14873ec2"
+dependencies = [
+ "twox-hash",
+]
+
 [[package]]
 name = "matchers"
 version = "0.0.1"
@@ -1180,6 +1189,7 @@ dependencies = [
  "hyper",
  "lazy_static",
  "log",
+ "lz4_flex",
  "postgres",
  "postgres-protocol",
  "postgres-types",
@@ -1198,6 +1208,7 @@ dependencies = [
  "toml",
  "tracing",
  "workspace_hack",
+ "yakv",
  "zenith_metrics",
  "zenith_utils",
 ]
@@ -1908,6 +1919,12 @@ version = "0.5.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "6e63cff320ae2c57904679ba7cb63280a3dc4613885beafb148ee7bf9aa9042d"
 
+[[package]]
+name = "static_assertions"
+version = "1.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a2eb9349b6444b326872e140eb1cf5e7c522154d69e7a0ffb0fb81c06b37543f"
+
 [[package]]
 name = "stringprep"
 version = "0.1.2"
@@ -2228,6 +2245,16 @@ version = "0.2.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "59547bce71d9c38b83d9c0e92b6066c4253371f15005def0c30d9657f50c7642"
 
+[[package]]
+name = "twox-hash"
+version = "1.6.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1f559b464de2e2bdabcac6a210d12e9b5a5973c251e102c44c585c71d51bd78e"
+dependencies = [
+ "cfg-if 1.0.0",
+ "static_assertions",
+]
+
 [[package]]
 name = "typenum"
 version = "1.13.0"
@@ -2542,6 +2569,14 @@ version = "0.8.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "b07db065a5cf61a7e4ba64f29e67db906fb1787316516c4e6e5ff0fea1efcd8a"
 
+[[package]]
+name = "yakv"
+version = "0.2.4"
+dependencies = [
+ "anyhow",
+ "fs2",
+]
+
 [[package]]
 name = "zenith"
 version = "0.1.0"

Cargo.toml

@@ -15,3 +15,7 @@ members = [
 # This is useful for profiling and, to some extent, debug.
 # Besides, debug info should not affect the performance.
 debug = true
+panic = 'abort'
+
+[profile.dev]
+panic = 'abort'

Dockerfile

@@ -38,8 +38,6 @@ RUN apt-get update && apt-get -yq install libreadline-dev libseccomp-dev openssl
 
 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/proxy /usr/local/bin
 COPY --from=pg-build /zenith/tmp_install postgres_install
 COPY docker-entrypoint.sh /docker-entrypoint.sh

Dockerfile.alpine

@@ -82,8 +82,6 @@ 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/proxy /usr/local/bin
 COPY --from=pg-build /zenith/tmp_install /usr/local
 COPY docker-entrypoint.sh /docker-entrypoint.sh

README.md

@@ -28,7 +28,7 @@ apt install build-essential libtool libreadline-dev zlib1g-dev flex bison libsec
 libssl-dev clang pkg-config libpq-dev
 ```
 
-[Rust] 1.52 or later is also required.
+[Rust] 1.55 or later is also required.
 
 To run the `psql` client, install the `postgresql-client` package or modify `PATH` and `LD_LIBRARY_PATH` to include `tmp_install/bin` and `tmp_install/lib`, respectively.
 

control_plane/src/compute.rs

@@ -84,25 +84,53 @@ impl ComputeControlPlane {
         }
     }
 
+    // FIXME: see also parse_point_in_time in branches.rs.
+    fn parse_point_in_time(
+        &self,
+        tenantid: ZTenantId,
+        s: &str,
+    ) -> Result<(ZTimelineId, Option<Lsn>)> {
+        let mut strings = s.split('@');
+        let name = strings.next().unwrap();
+
+        let lsn: Option<Lsn>;
+        if let Some(lsnstr) = strings.next() {
+            lsn = Some(
+                Lsn::from_str(lsnstr)
+                    .with_context(|| "invalid LSN in point-in-time specification")?,
+            );
+        } else {
+            lsn = None
+        }
+
+        // Resolve the timeline ID, given the human-readable branch name
+        let timeline_id = self
+            .pageserver
+            .branch_get_by_name(&tenantid, name)?
+            .timeline_id;
+
+        Ok((timeline_id, lsn))
+    }
+
     pub fn new_node(
         &mut self,
         tenantid: ZTenantId,
-        branch_name: &str,
+        name: &str,
+        timeline_spec: &str,
         port: Option<u16>,
     ) -> Result<Arc<PostgresNode>> {
-        let timeline_id = self
-            .pageserver
-            .branch_get_by_name(&tenantid, branch_name)?
-            .timeline_id;
+        // Resolve the human-readable timeline spec into timeline ID and LSN
+        let (timelineid, lsn) = self.parse_point_in_time(tenantid, timeline_spec)?;
 
         let port = port.unwrap_or_else(|| self.get_port());
         let node = Arc::new(PostgresNode {
-            name: branch_name.to_owned(),
+            name: name.to_owned(),
             address: SocketAddr::new("127.0.0.1".parse().unwrap(), port),
             env: self.env.clone(),
             pageserver: Arc::clone(&self.pageserver),
             is_test: false,
-            timelineid: timeline_id,
+            timelineid,
+            lsn,
             tenantid,
             uses_wal_proposer: false,
         });
@@ -127,6 +155,7 @@ pub struct PostgresNode {
     pageserver: Arc<PageServerNode>,
     is_test: bool,
     pub timelineid: ZTimelineId,
+    pub lsn: Option<Lsn>, // if it's a read-only node. None for primary
     pub tenantid: ZTenantId,
     uses_wal_proposer: bool,
 }
@@ -161,9 +190,12 @@ impl PostgresNode {
         let port: u16 = conf.parse_field("port", &context)?;
         let timelineid: ZTimelineId = conf.parse_field("zenith.zenith_timeline", &context)?;
         let tenantid: ZTenantId = conf.parse_field("zenith.zenith_tenant", &context)?;
-
         let uses_wal_proposer = conf.get("wal_acceptors").is_some();
 
+        // parse recovery_target_lsn, if any
+        let recovery_target_lsn: Option<Lsn> =
+            conf.parse_field_optional("recovery_target_lsn", &context)?;
+
         // ok now
         Ok(PostgresNode {
             address: SocketAddr::new("127.0.0.1".parse().unwrap(), port),
@@ -172,6 +204,7 @@ impl PostgresNode {
             pageserver: Arc::clone(pageserver),
             is_test: false,
             timelineid,
+            lsn: recovery_target_lsn,
             tenantid,
             uses_wal_proposer,
         })
@@ -233,7 +266,7 @@ impl PostgresNode {
         // Read the archive directly from the `CopyOutReader`
         tar::Archive::new(copyreader)
             .unpack(&self.pgdata())
-            .with_context(|| "extracting page backup failed")?;
+            .with_context(|| "extracting base backup failed")?;
 
         Ok(())
     }
@@ -265,6 +298,7 @@ impl PostgresNode {
         conf.append("max_replication_slots", "10");
         conf.append("hot_standby", "on");
         conf.append("shared_buffers", "1MB");
+        conf.append("max_wal_size", "100GB");
         conf.append("fsync", "off");
         conf.append("max_connections", "100");
         conf.append("wal_sender_timeout", "0");
@@ -301,6 +335,9 @@ impl PostgresNode {
         conf.append("zenith.page_server_connstring", &pageserver_connstr);
         conf.append("zenith.zenith_tenant", &self.tenantid.to_string());
         conf.append("zenith.zenith_timeline", &self.timelineid.to_string());
+        if let Some(lsn) = self.lsn {
+            conf.append("recovery_target_lsn", &lsn.to_string());
+        }
         conf.append_line("");
 
         // Configure the node to stream WAL directly to the pageserver
@@ -314,7 +351,9 @@ impl PostgresNode {
     }
 
     fn load_basebackup(&self) -> Result<()> {
-        let lsn = if self.uses_wal_proposer {
+        let backup_lsn = if let Some(lsn) = self.lsn {
+            Some(lsn)
+        } else if self.uses_wal_proposer {
             // LSN 0 means that it is bootstrap and we need to download just
             // latest data from the pageserver. That is a bit clumsy but whole bootstrap
             // procedure evolves quite actively right now, so let's think about it again
@@ -329,7 +368,7 @@ impl PostgresNode {
             None
         };
 
-        self.do_basebackup(lsn)?;
+        self.do_basebackup(backup_lsn)?;
 
         Ok(())
     }
@@ -406,6 +445,10 @@ impl PostgresNode {
         // 3. Load basebackup
         self.load_basebackup()?;
 
+        if self.lsn.is_some() {
+            File::create(self.pgdata().join("standby.signal"))?;
+        }
+
         // 4. Finally start the compute node postgres
         println!("Starting postgres node at '{}'", self.connstr());
         self.pg_ctl(&["start"], auth_token)

control_plane/src/postgresql_conf.rs

@@ -83,6 +83,22 @@ impl PostgresConf {
             .with_context(|| format!("could not parse '{}' option {}", field_name, context))
     }
 
+    pub fn parse_field_optional<T>(&self, field_name: &str, context: &str) -> Result<Option<T>>
+    where
+        T: FromStr,
+        <T as FromStr>::Err: std::error::Error + Send + Sync + 'static,
+    {
+        if let Some(val) = self.get(field_name) {
+            let result = val
+                .parse::<T>()
+                .with_context(|| format!("could not parse '{}' option {}", field_name, context))?;
+
+            Ok(Some(result))
+        } else {
+            Ok(None)
+        }
+    }
+
     ///
     /// Note: if you call this multiple times for the same option, the config
     /// file will a line for each call. It would be nice to have a function

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",

pageserver/Cargo.toml

@@ -37,6 +37,9 @@ async-trait = "0.1"
 const_format = "0.2.21"
 tracing = "0.1.27"
 signal-hook = {version = "0.3.10", features = ["extended-siginfo"] }
+#yakv = { path  = "../../yakv" }
+yakv = "0.2.4"
+lz4_flex = "0.9.0"
 
 postgres_ffi = { path = "../postgres_ffi" }
 zenith_metrics = { path = "../zenith_metrics" }

pageserver/src/basebackup.rs

@@ -13,6 +13,7 @@
 use anyhow::Result;
 use bytes::{BufMut, BytesMut};
 use log::*;
+use std::fmt::Write as FmtWrite;
 use std::io;
 use std::io::Write;
 use std::sync::Arc;
@@ -83,7 +84,7 @@ impl<'a> Basebackup<'a> {
 
         info!(
             "taking basebackup lsn={}, prev_lsn={}",
-            backup_prev, backup_lsn
+            backup_lsn, backup_prev
         );
 
         Ok(Basebackup {
@@ -248,13 +249,7 @@ impl<'a> Basebackup<'a> {
         let mut pg_control = ControlFileData::decode(&pg_control_bytes)?;
         let mut checkpoint = CheckPoint::decode(&checkpoint_bytes)?;
 
-        // Generate new pg_control and WAL needed for bootstrap
-        let checkpoint_segno = self.lsn.segment_number(pg_constants::WAL_SEGMENT_SIZE);
-        let checkpoint_lsn = XLogSegNoOffsetToRecPtr(
-            checkpoint_segno,
-            XLOG_SIZE_OF_XLOG_LONG_PHD as u32,
-            pg_constants::WAL_SEGMENT_SIZE,
-        );
+        // Generate new pg_control needed for bootstrap
         checkpoint.redo = normalize_lsn(self.lsn, pg_constants::WAL_SEGMENT_SIZE).0;
 
         //reset some fields we don't want to preserve
@@ -263,19 +258,24 @@ impl<'a> Basebackup<'a> {
         checkpoint.oldestActiveXid = 0;
 
         //save new values in pg_control
-        pg_control.checkPoint = checkpoint_lsn;
+        pg_control.checkPoint = 0;
         pg_control.checkPointCopy = checkpoint;
         pg_control.state = pg_constants::DB_SHUTDOWNED;
 
         // add zenith.signal file
-        let xl_prev = if self.prev_record_lsn == Lsn(0) {
-            0xBAD0 // magic value to indicate that we don't know prev_lsn
+        let mut zenith_signal = String::new();
+        if self.prev_record_lsn == Lsn(0) {
+            if self.lsn == self.timeline.get_ancestor_lsn() {
+                write!(zenith_signal, "PREV LSN: none")?;
+            } else {
+                write!(zenith_signal, "PREV LSN: invalid")?;
+            }
         } else {
-            self.prev_record_lsn.0
-        };
+            write!(zenith_signal, "PREV LSN: {}", self.prev_record_lsn)?;
+        }
         self.ar.append(
-            &new_tar_header("zenith.signal", 8)?,
-            &xl_prev.to_le_bytes()[..],
+            &new_tar_header("zenith.signal", zenith_signal.len() as u64)?,
+            zenith_signal.as_bytes(),
         )?;
 
         //send pg_control
@@ -284,14 +284,15 @@ impl<'a> Basebackup<'a> {
         self.ar.append(&header, &pg_control_bytes[..])?;
 
         //send wal segment
+        let segno = self.lsn.segment_number(pg_constants::WAL_SEGMENT_SIZE);
         let wal_file_name = XLogFileName(
             1, // FIXME: always use Postgres timeline 1
-            checkpoint_segno,
+            segno,
             pg_constants::WAL_SEGMENT_SIZE,
         );
         let wal_file_path = format!("pg_wal/{}", wal_file_name);
         let header = new_tar_header(&wal_file_path, pg_constants::WAL_SEGMENT_SIZE as u64)?;
-        let wal_seg = generate_wal_segment(&pg_control);
+        let wal_seg = generate_wal_segment(segno, pg_control.system_identifier);
         assert!(wal_seg.len() == pg_constants::WAL_SEGMENT_SIZE);
         self.ar.append(&header, &wal_seg[..])?;
         Ok(())

pageserver/src/bin/dump_layerfile.rs

@@ -1,25 +0,0 @@
-//! Main entry point for the dump_layerfile executable
-//!
-//! A handy tool for debugging, that's all.
-use anyhow::Result;
-use clap::{App, Arg};
-use pageserver::layered_repository::dump_layerfile_from_path;
-use std::path::PathBuf;
-
-fn main() -> Result<()> {
-    let arg_matches = App::new("Zenith dump_layerfile utility")
-        .about("Dump contents of one layer file, for debugging")
-        .arg(
-            Arg::with_name("path")
-                .help("Path to file to dump")
-                .required(true)
-                .index(1),
-        )
-        .get_matches();
-
-    let path = PathBuf::from(arg_matches.value_of("path").unwrap());
-
-    dump_layerfile_from_path(&path)?;
-
-    Ok(())
-}

pageserver/src/bin/pageserver.rs

@@ -2,7 +2,6 @@
 // Main entry point for the Page Server executable
 //
 
-use pageserver::defaults::*;
 use serde::{Deserialize, Serialize};
 use std::{
     env,
@@ -28,13 +27,8 @@ use clap::{App, Arg, ArgMatches};
 use daemonize::Daemonize;
 
 use pageserver::{
-    branches,
-    defaults::{
-        DEFAULT_HTTP_LISTEN_ADDR, DEFAULT_PG_LISTEN_ADDR,
-        DEFAULT_RELISH_STORAGE_MAX_CONCURRENT_SYNC_LIMITS,
-    },
-    http, page_service, relish_storage, tenant_mgr, PageServerConf, RelishStorageConfig,
-    RelishStorageKind, S3Config, LOG_FILE_NAME,
+    branches, defaults::*, http, page_service, relish_storage, tenant_mgr, PageServerConf,
+    RelishStorageConfig, RelishStorageKind, S3Config, LOG_FILE_NAME,
 };
 use zenith_utils::http::endpoint;
 use zenith_utils::postgres_backend;
@@ -42,23 +36,35 @@ use zenith_utils::postgres_backend;
 use const_format::formatcp;
 
 /// String arguments that can be declared via CLI or config file
-#[derive(Serialize, Deserialize)]
+#[derive(Serialize, Deserialize, PartialEq, Eq, Clone)]
 struct CfgFileParams {
     listen_pg_addr: Option<String>,
     listen_http_addr: Option<String>,
     checkpoint_distance: Option<String>,
     checkpoint_period: Option<String>,
+    upload_distance: Option<String>,
+    upload_period: Option<String>,
+    reconstruct_threshold: Option<String>,
     gc_horizon: Option<String>,
     gc_period: Option<String>,
     pg_distrib_dir: Option<String>,
     auth_validation_public_key_path: Option<String>,
     auth_type: Option<String>,
-    // see https://github.com/alexcrichton/toml-rs/blob/6c162e6562c3e432bf04c82a3d1d789d80761a86/examples/enum_external.rs for enum deserialisation examples
-    relish_storage: Option<RelishStorage>,
     relish_storage_max_concurrent_sync: Option<String>,
+    /////////////////////////////////
+    //// Don't put `Option<String>` and other "simple" values below.
+    ////
+    /// `Option<RelishStorage>` is a <a href='https://toml.io/en/v1.0.0#table'>table</a> in TOML.
+    /// Values in TOML cannot be defined after tables (other tables can),
+    /// and [`toml`] crate serializes all fields in the order of their appearance.
+    ////////////////////////////////
+    relish_storage: Option<RelishStorage>,
 }
 
-#[derive(Serialize, Deserialize, Clone)]
+#[derive(Serialize, Deserialize, PartialEq, Eq, Clone)]
+// Without this attribute, enums with values won't be serialized by the `toml` library (but can be deserialized nonetheless!).
+// See https://github.com/alexcrichton/toml-rs/blob/6c162e6562c3e432bf04c82a3d1d789d80761a86/examples/enum_external.rs for the examples
+#[serde(untagged)]
 enum RelishStorage {
     Local {
         local_path: String,
@@ -100,6 +106,9 @@ impl CfgFileParams {
             listen_http_addr: get_arg("listen-http"),
             checkpoint_distance: get_arg("checkpoint_distance"),
             checkpoint_period: get_arg("checkpoint_period"),
+            upload_distance: get_arg("upload_distance"),
+            upload_period: get_arg("upload_period"),
+            reconstruct_threshold: get_arg("reconstruct_threshold"),
             gc_horizon: get_arg("gc_horizon"),
             gc_period: get_arg("gc_period"),
             pg_distrib_dir: get_arg("postgres-distrib"),
@@ -118,6 +127,9 @@ impl CfgFileParams {
             listen_http_addr: self.listen_http_addr.or(other.listen_http_addr),
             checkpoint_distance: self.checkpoint_distance.or(other.checkpoint_distance),
             checkpoint_period: self.checkpoint_period.or(other.checkpoint_period),
+            upload_distance: self.upload_distance.or(other.upload_distance),
+            upload_period: self.upload_period.or(other.upload_period),
+            reconstruct_threshold: self.reconstruct_threshold.or(other.reconstruct_threshold),
             gc_horizon: self.gc_horizon.or(other.gc_horizon),
             gc_period: self.gc_period.or(other.gc_period),
             pg_distrib_dir: self.pg_distrib_dir.or(other.pg_distrib_dir),
@@ -155,6 +167,20 @@ impl CfgFileParams {
             None => DEFAULT_CHECKPOINT_PERIOD,
         };
 
+        let upload_distance: u64 = match self.upload_distance.as_ref() {
+            Some(upload_distance_str) => upload_distance_str.parse()?,
+            None => DEFAULT_UPLOAD_DISTANCE,
+        };
+        let upload_period = match self.upload_period.as_ref() {
+            Some(upload_period_str) => humantime::parse_duration(upload_period_str)?,
+            None => DEFAULT_UPLOAD_PERIOD,
+        };
+
+        let reconstruct_threshold: u64 = match self.reconstruct_threshold.as_ref() {
+            Some(reconstruct_threshold_str) => reconstruct_threshold_str.parse()?,
+            None => DEFAULT_RECONSTRUCT_THRESHOLD,
+        };
+
         let gc_horizon: u64 = match self.gc_horizon.as_ref() {
             Some(horizon_str) => horizon_str.parse()?,
             None => DEFAULT_GC_HORIZON,
@@ -233,6 +259,9 @@ impl CfgFileParams {
             listen_http_addr,
             checkpoint_distance,
             checkpoint_period,
+            upload_distance,
+            upload_period,
+            reconstruct_threshold,
             gc_horizon,
             gc_period,
 
@@ -293,6 +322,24 @@ fn main() -> Result<()> {
                 .takes_value(true)
                 .help("Interval between checkpoint iterations"),
         )
+        .arg(
+            Arg::with_name("checkpoint_distance")
+                .long("checkpoint_distance")
+                .takes_value(true)
+                .help("Distance from current LSN to perform checkpoint of in-memory layers"),
+        )
+        .arg(
+            Arg::with_name("upload_period")
+                .long("upload_period")
+                .takes_value(true)
+                .help("Interval between upload iterations"),
+        )
+        .arg(
+            Arg::with_name("reconstruct_threshold")
+                .long("reconstruct_threshold")
+                .takes_value(true)
+                .help("Minimal size of deltas after which page reconstruction (materialization) can be performed"),
+        )
         .arg(
             Arg::with_name("gc_horizon")
                 .long("gc_horizon")
@@ -552,7 +599,7 @@ fn start_pageserver(conf: &'static PageServerConf) -> Result<()> {
                 info!("Got SIGQUIT. Terminate pageserver in immediate shutdown mode");
                 exit(111);
             }
-            SIGTERM => {
+            SIGINT | SIGTERM => {
                 info!("Got SIGINT/SIGTERM. Terminate gracefully in fast shutdown mode");
                 // Terminate postgres backends
                 postgres_backend::set_pgbackend_shutdown_requested();
@@ -577,11 +624,154 @@ fn start_pageserver(conf: &'static PageServerConf) -> Result<()> {
                 info!("Pageserver shut down successfully completed");
                 exit(0);
             }
-            _ => {
-                debug!("Unknown signal.");
+            unknown_signal => {
+                debug!("Unknown signal {}", unknown_signal);
             }
         }
     }
 
     Ok(())
 }
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn page_server_conf_toml_serde() {
+        let params = CfgFileParams {
+            listen_pg_addr: Some("listen_pg_addr_VALUE".to_string()),
+            listen_http_addr: Some("listen_http_addr_VALUE".to_string()),
+            checkpoint_distance: Some("checkpoint_distance_VALUE".to_string()),
+            checkpoint_period: Some("checkpoint_period_VALUE".to_string()),
+            upload_distance: Some("upload_distance_VALUE".to_string()),
+            upload_period: Some("upload_period_VALUE".to_string()),
+            reconstruct_threshold: Some("reconstruct_threshold_VALUE".to_string()),
+            gc_horizon: Some("gc_horizon_VALUE".to_string()),
+            gc_period: Some("gc_period_VALUE".to_string()),
+            pg_distrib_dir: Some("pg_distrib_dir_VALUE".to_string()),
+            auth_validation_public_key_path: Some(
+                "auth_validation_public_key_path_VALUE".to_string(),
+            ),
+            auth_type: Some("auth_type_VALUE".to_string()),
+            relish_storage: Some(RelishStorage::Local {
+                local_path: "relish_storage_local_VALUE".to_string(),
+            }),
+            relish_storage_max_concurrent_sync: Some(
+                "relish_storage_max_concurrent_sync_VALUE".to_string(),
+            ),
+        };
+
+        let toml_string = toml::to_string(&params).expect("Failed to serialize correct config");
+        let toml_pretty_string =
+            toml::to_string_pretty(&params).expect("Failed to serialize correct config");
+        assert_eq!(
+            r#"listen_pg_addr = 'listen_pg_addr_VALUE'
+listen_http_addr = 'listen_http_addr_VALUE'
+checkpoint_distance = 'checkpoint_distance_VALUE'
+checkpoint_period = 'checkpoint_period_VALUE'
+upload_distance = 'upload_distance_VALUE'
+upload_period = 'upload_period_VALUE'
+reconstruct_threshold = 'reconstruct_threshold_VALUE'
+gc_horizon = 'gc_horizon_VALUE'
+gc_period = 'gc_period_VALUE'
+pg_distrib_dir = 'pg_distrib_dir_VALUE'
+auth_validation_public_key_path = 'auth_validation_public_key_path_VALUE'
+auth_type = 'auth_type_VALUE'
+relish_storage_max_concurrent_sync = 'relish_storage_max_concurrent_sync_VALUE'
+
+[relish_storage]
+local_path = 'relish_storage_local_VALUE'
+"#,
+            toml_pretty_string
+        );
+
+        let params_from_serialized: CfgFileParams = toml::from_str(&toml_string)
+            .expect("Failed to deserialize the serialization result of the config");
+        let params_from_serialized_pretty: CfgFileParams = toml::from_str(&toml_pretty_string)
+            .expect("Failed to deserialize the prettified serialization result of the config");
+        assert!(
+            params_from_serialized == params,
+            "Expected the same config in the end of config -> serialize -> deserialize chain"
+        );
+        assert!(
+            params_from_serialized_pretty == params,
+            "Expected the same config in the end of config -> serialize pretty -> deserialize chain"
+        );
+    }
+
+    #[test]
+    fn credentials_omitted_during_serialization() {
+        let params = CfgFileParams {
+            listen_pg_addr: Some("listen_pg_addr_VALUE".to_string()),
+            listen_http_addr: Some("listen_http_addr_VALUE".to_string()),
+            checkpoint_distance: Some("checkpoint_distance_VALUE".to_string()),
+            checkpoint_period: Some("checkpoint_period_VALUE".to_string()),
+            upload_distance: Some("upload_distance_VALUE".to_string()),
+            upload_period: Some("upload_period_VALUE".to_string()),
+            reconstruct_threshold: Some("reconstruct_threshold_VALUE".to_string()),
+            gc_horizon: Some("gc_horizon_VALUE".to_string()),
+            gc_period: Some("gc_period_VALUE".to_string()),
+            pg_distrib_dir: Some("pg_distrib_dir_VALUE".to_string()),
+            auth_validation_public_key_path: Some(
+                "auth_validation_public_key_path_VALUE".to_string(),
+            ),
+            auth_type: Some("auth_type_VALUE".to_string()),
+            relish_storage: Some(RelishStorage::AwsS3 {
+                bucket_name: "bucket_name_VALUE".to_string(),
+                bucket_region: "bucket_region_VALUE".to_string(),
+                access_key_id: Some("access_key_id_VALUE".to_string()),
+                secret_access_key: Some("secret_access_key_VALUE".to_string()),
+            }),
+            relish_storage_max_concurrent_sync: Some(
+                "relish_storage_max_concurrent_sync_VALUE".to_string(),
+            ),
+        };
+
+        let toml_string = toml::to_string(&params).expect("Failed to serialize correct config");
+        let toml_pretty_string =
+            toml::to_string_pretty(&params).expect("Failed to serialize correct config");
+        assert_eq!(
+            r#"listen_pg_addr = 'listen_pg_addr_VALUE'
+listen_http_addr = 'listen_http_addr_VALUE'
+checkpoint_distance = 'checkpoint_distance_VALUE'
+checkpoint_period = 'checkpoint_period_VALUE'
+upload_distance = 'upload_distance_VALUE'
+upload_period = 'upload_period_VALUE'
+reconstruct_threshold = 'reconstruct_threshold_VALUE'
+gc_horizon = 'gc_horizon_VALUE'
+gc_period = 'gc_period_VALUE'
+pg_distrib_dir = 'pg_distrib_dir_VALUE'
+auth_validation_public_key_path = 'auth_validation_public_key_path_VALUE'
+auth_type = 'auth_type_VALUE'
+relish_storage_max_concurrent_sync = 'relish_storage_max_concurrent_sync_VALUE'
+
+[relish_storage]
+bucket_name = 'bucket_name_VALUE'
+bucket_region = 'bucket_region_VALUE'
+"#,
+            toml_pretty_string
+        );
+
+        let params_from_serialized: CfgFileParams = toml::from_str(&toml_string)
+            .expect("Failed to deserialize the serialization result of the config");
+        let params_from_serialized_pretty: CfgFileParams = toml::from_str(&toml_pretty_string)
+            .expect("Failed to deserialize the prettified serialization result of the config");
+
+        let mut expected_params = params;
+        expected_params.relish_storage = Some(RelishStorage::AwsS3 {
+            bucket_name: "bucket_name_VALUE".to_string(),
+            bucket_region: "bucket_region_VALUE".to_string(),
+            access_key_id: None,
+            secret_access_key: None,
+        });
+        assert!(
+            params_from_serialized == expected_params,
+            "Expected the config without credentials in the end of a 'config -> serialize -> deserialize' chain"
+        );
+        assert!(
+            params_from_serialized_pretty == expected_params,
+            "Expected the config without credentials in the end of a 'config -> serialize pretty -> deserialize' chain"
+        );
+    }
+}

pageserver/src/branches.rs

@@ -147,7 +147,7 @@ pub fn create_repo(
 
     let tli = create_timeline(conf, None, &tenantid)?;
 
-    let repo = Arc::new(crate::layered_repository::LayeredRepository::new(
+    let repo = Arc::new(crate::buffered_repository::BufferedRepository::new(
         conf,
         wal_redo_manager,
         tenantid,

pageserver/src/layered_repository/delta_layer.rs

@@ -39,9 +39,8 @@
 //!
 use crate::layered_repository::blob::BlobWriter;
 use crate::layered_repository::filename::{DeltaFileName, PathOrConf};
-use crate::layered_repository::storage_layer::{
-    Layer, PageReconstructData, PageReconstructResult, PageVersion, SegmentTag,
-};
+use crate::layered_repository::storage_layer::{Layer, SegmentTag};
+use crate::repository::{PageReconstructData, PageReconstructResult, PageVersion};
 use crate::waldecoder;
 use crate::PageServerConf;
 use crate::{ZTenantId, ZTimelineId};
@@ -148,6 +147,10 @@ pub struct DeltaLayerInner {
 }
 
 impl Layer for DeltaLayer {
+    fn get_tenant_id(&self) -> ZTenantId {
+        self.tenantid
+    }
+
     fn get_timeline_id(&self) -> ZTimelineId {
         self.timelineid
     }
@@ -201,22 +204,22 @@ impl Layer for DeltaLayer {
             for ((_blknum, pv_lsn), blob_range) in iter {
                 let pv = PageVersion::des(&read_blob(&page_version_reader, blob_range)?)?;
 
-                if let Some(img) = pv.page_image {
-                    // Found a page image, return it
-                    reconstruct_data.page_img = Some(img);
-                    need_image = false;
-                    break;
-                } else if let Some(rec) = pv.record {
-                    let will_init = rec.will_init;
-                    reconstruct_data.records.push((*pv_lsn, rec));
-                    if will_init {
-                        // This WAL record initializes the page, so no need to go further back
+                match pv {
+                    PageVersion::Page(img) => {
+                        // Found a page image, return it
+                        reconstruct_data.page_img = Some(img);
                         need_image = false;
                         break;
                     }
-                } else {
-                    // No base image, and no WAL record. Huh?
-                    bail!("no page image or WAL record for requested page");
+                    PageVersion::Wal(rec) => {
+                        let will_init = rec.will_init;
+                        reconstruct_data.records.push((*pv_lsn, rec));
+                        if will_init {
+                            // This WAL record initializes the page, so no need to go further back
+                            need_image = false;
+                            break;
+                        }
+                    }
                 }
             }
 
@@ -226,7 +229,7 @@ impl Layer for DeltaLayer {
         // If an older page image is needed to reconstruct the page, let the
         // caller know.
         if need_image {
-            Ok(PageReconstructResult::Continue(self.start_lsn))
+            Ok(PageReconstructResult::Continue(Lsn(self.start_lsn.0 - 1)))
         } else {
             Ok(PageReconstructResult::Complete)
         }
@@ -307,19 +310,22 @@ impl Layer for DeltaLayer {
             let buf = read_blob(&chapter, blob_range)?;
             let pv = PageVersion::des(&buf)?;
 
-            if let Some(img) = pv.page_image.as_ref() {
-                write!(&mut desc, " img {} bytes", img.len())?;
-            }
-            if let Some(rec) = pv.record.as_ref() {
-                let wal_desc = waldecoder::describe_wal_record(&rec.rec);
-                write!(
-                    &mut desc,
-                    " rec {} bytes will_init: {} {}",
-                    rec.rec.len(),
-                    rec.will_init,
-                    wal_desc
-                )?;
+            match pv {
+                PageVersion::Page(img) => {
+                    write!(&mut desc, " img {} bytes", img.len())?;
+                }
+                PageVersion::Wal(rec) => {
+                    let wal_desc = waldecoder::describe_wal_record(&rec.rec);
+                    write!(
+                        &mut desc,
+                        " rec {} bytes will_init: {} {}",
+                        rec.rec.len(),
+                        rec.will_init,
+                        wal_desc
+                    )?;
+                }
             }
+
             println!("  blk {} at {}: {}", blk, lsn, desc);
         }
 
@@ -328,6 +334,19 @@ impl Layer for DeltaLayer {
 }
 
 impl DeltaLayer {
+    /// debugging function to print out the contents of the layer
+    pub fn versions(&self) -> Result<Vec<(u32, Lsn, PageVersion)>> {
+        let mut versions: Vec<(u32, Lsn, PageVersion)> = Vec::new();
+        let inner = self.load()?;
+        let (_path, book) = self.open_book()?;
+        let chapter = book.chapter_reader(PAGE_VERSIONS_CHAPTER)?;
+        for ((blk, lsn), blob_range) in inner.page_version_metas.as_slice() {
+            let buf = read_blob(&chapter, blob_range)?;
+            versions.push((*blk, *lsn, PageVersion::des(&buf)?));
+        }
+        Ok(versions)
+    }
+
     fn path_for(
         path_or_conf: &PathOrConf,
         timelineid: ZTimelineId,
@@ -360,6 +379,7 @@ impl DeltaLayer {
         dropped: bool,
         page_versions: impl Iterator<Item = (u32, Lsn, &'a PageVersion)>,
         relsizes: VecMap<Lsn, u32>,
+        nosync: bool,
     ) -> Result<DeltaLayer> {
         if seg.rel.is_blocky() {
             assert!(!relsizes.is_empty());
@@ -432,8 +452,9 @@ impl DeltaLayer {
 
         // This flushes the underlying 'buf_writer'.
         let writer = book.close()?;
-        writer.get_ref().sync_all()?;
-
+        if !nosync {
+            writer.get_ref().sync_all()?;
+        }
         trace!("saved {}", &path.display());
 
         drop(inner);
@@ -442,12 +463,7 @@ impl DeltaLayer {
     }
 
     fn open_book(&self) -> Result<(PathBuf, Book<File>)> {
-        let path = Self::path_for(
-            &self.path_or_conf,
-            self.timelineid,
-            self.tenantid,
-            &self.layer_name(),
-        );
+        let path = self.path();
 
         let file = File::open(&path)?;
         let book = Book::new(file)?;

pageserver/src/layered_repository/filename.rs

@@ -13,8 +13,6 @@ use anyhow::Result;
 use log::*;
 use zenith_utils::lsn::Lsn;
 
-use super::METADATA_FILE_NAME;
-
 // Note: LayeredTimeline::load_layer_map() relies on this sort order
 #[derive(Debug, PartialEq, Eq, PartialOrd, Ord, Clone)]
 pub struct DeltaFileName {
@@ -292,7 +290,7 @@ pub fn list_files(
             deltafiles.push(deltafilename);
         } else if let Some(imgfilename) = ImageFileName::parse_str(fname) {
             imgfiles.push(imgfilename);
-        } else if fname == METADATA_FILE_NAME || fname == "ancestor" || fname.ends_with(".old") {
+        } else if fname == "metadata" || fname == "ancestor" || fname.ends_with(".old") {
             // ignore these
         } else {
             warn!("unrecognized filename in timeline dir: {}", fname);

pageserver/src/layered_repository/image_layer.rs

@@ -22,11 +22,8 @@
 //! For non-blocky relishes, the image can be found in NONBLOCKY_IMAGE_CHAPTER.
 //!
 use crate::layered_repository::filename::{ImageFileName, PathOrConf};
-use crate::layered_repository::storage_layer::{
-    Layer, PageReconstructData, PageReconstructResult, SegmentTag,
-};
-use crate::layered_repository::LayeredTimeline;
-use crate::layered_repository::RELISH_SEG_SIZE;
+use crate::layered_repository::storage_layer::{Layer, SegmentTag, RELISH_SEG_SIZE};
+use crate::repository::{PageReconstructData, PageReconstructResult};
 use crate::PageServerConf;
 use crate::{ZTenantId, ZTimelineId};
 use anyhow::{anyhow, bail, ensure, Result};
@@ -117,6 +114,10 @@ impl Layer for ImageLayer {
         PathBuf::from(self.layer_name().to_string())
     }
 
+    fn get_tenant_id(&self) -> ZTenantId {
+        self.tenantid
+    }
+
     fn get_timeline_id(&self) -> ZTimelineId {
         self.timelineid
     }
@@ -250,13 +251,14 @@ impl ImageLayer {
     }
 
     /// Create a new image file, using the given array of pages.
-    fn create(
+    pub fn create(
         conf: &'static PageServerConf,
         timelineid: ZTimelineId,
         tenantid: ZTenantId,
         seg: SegmentTag,
         lsn: Lsn,
         base_images: Vec<Bytes>,
+        nosync: bool,
     ) -> Result<ImageLayer> {
         let image_type = if seg.rel.is_blocky() {
             let num_blocks: u32 = base_images.len().try_into()?;
@@ -316,8 +318,9 @@ impl ImageLayer {
 
         // This flushes the underlying 'buf_writer'.
         let writer = book.close()?;
-        writer.get_ref().sync_all()?;
-
+        if !nosync {
+            writer.get_ref().sync_all()?;
+        }
         trace!("saved {}", path.display());
 
         drop(inner);
@@ -325,6 +328,7 @@ impl ImageLayer {
         Ok(layer)
     }
 
+    /*
     // Create a new image file by materializing every page in a source layer
     // at given LSN.
     pub fn create_from_src(
@@ -362,6 +366,7 @@ impl ImageLayer {
 
         Self::create(conf, timelineid, timeline.tenantid, seg, lsn, base_images)
     }
+    */
 
     ///
     /// Load the contents of the file into memory

pageserver/src/layered_repository/storage_layer.rs

@@ -3,10 +3,9 @@
 //!
 
 use crate::relish::RelishTag;
-use crate::repository::WALRecord;
-use crate::ZTimelineId;
+use crate::repository::{PageReconstructData, PageReconstructResult};
+use crate::{ZTenantId, ZTimelineId};
 use anyhow::Result;
-use bytes::Bytes;
 use serde::{Deserialize, Serialize};
 use std::fmt;
 use std::path::PathBuf;
@@ -45,56 +44,6 @@ impl SegmentTag {
     }
 }
 
-///
-/// Represents a version of a page at a specific LSN. The LSN is the key of the
-/// entry in the 'page_versions' hash, it is not duplicated here.
-///
-/// A page version can be stored as a full page image, or as WAL record that needs
-/// to be applied over the previous page version to reconstruct this version.
-///
-/// It's also possible to have both a WAL record and a page image in the same
-/// PageVersion. That happens if page version is originally stored as a WAL record
-/// but it is later reconstructed by a GetPage@LSN request by performing WAL
-/// redo. The get_page_at_lsn() code will store the reconstructed pag image next to
-/// the WAL record in that case. TODO: That's pretty accidental, not the result
-/// of any grand design. If we want to keep reconstructed page versions around, we
-/// probably should have a separate buffer cache so that we could control the
-/// replacement policy globally. Or if we keep a reconstructed page image, we
-/// could throw away the WAL record.
-///
-#[derive(Debug, Clone, Serialize, Deserialize)]
-pub struct PageVersion {
-    /// an 8kb page image
-    pub page_image: Option<Bytes>,
-    /// WAL record to get from previous page version to this one.
-    pub record: Option<WALRecord>,
-}
-
-///
-/// Data needed to reconstruct a page version
-///
-/// 'page_img' is the old base image of the page to start the WAL replay with.
-/// It can be None, if the first WAL record initializes the page (will_init)
-/// 'records' contains the records to apply over the base image.
-///
-pub struct PageReconstructData {
-    pub records: Vec<(Lsn, WALRecord)>,
-    pub page_img: Option<Bytes>,
-}
-
-/// Return value from Layer::get_page_reconstruct_data
-pub enum PageReconstructResult {
-    /// Got all the data needed to reconstruct the requested page
-    Complete,
-    /// This layer didn't contain all the required data, the caller should look up
-    /// the predecessor layer at the returned LSN and collect more data from there.
-    Continue(Lsn),
-    /// This layer didn't contain data needed to reconstruct the page version at
-    /// the returned LSN. This is usually considered an error, but might be OK
-    /// in some circumstances.
-    Missing(Lsn),
-}
-
 ///
 /// A Layer corresponds to one RELISH_SEG_SIZE slice of a relish in a range of LSNs.
 /// There are two kinds of layers, in-memory and on-disk layers. In-memory
@@ -104,6 +53,8 @@ pub enum PageReconstructResult {
 /// in-memory and on-disk layers.
 ///
 pub trait Layer: Send + Sync {
+    fn get_tenant_id(&self) -> ZTenantId;
+
     /// Identify the timeline this relish belongs to
     fn get_timeline_id(&self) -> ZTimelineId;
 

pageserver/src/lib.rs

@@ -9,6 +9,7 @@ use zenith_metrics::{register_int_gauge_vec, IntGaugeVec};
 
 pub mod basebackup;
 pub mod branches;
+pub mod buffered_repository;
 pub mod http;
 pub mod layered_repository;
 pub mod page_service;
@@ -17,6 +18,7 @@ pub mod relish_storage;
 pub mod repository;
 pub mod restore_local_repo;
 pub mod tenant_mgr;
+pub mod toast_store;
 pub mod waldecoder;
 pub mod walreceiver;
 pub mod walredo;
@@ -30,14 +32,17 @@ pub mod defaults {
     pub const DEFAULT_HTTP_LISTEN_PORT: u16 = 9898;
     pub const DEFAULT_HTTP_LISTEN_ADDR: &str = formatcp!("127.0.0.1:{DEFAULT_HTTP_LISTEN_PORT}");
 
-    // FIXME: This current value is very low. I would imagine something like 1 GB or 10 GB
-    // would be more appropriate. But a low value forces the code to be exercised more,
-    // which is good for now to trigger bugs.
+    // Minimal size of WAL records chain to trigger materialization of the page
     pub const DEFAULT_CHECKPOINT_DISTANCE: u64 = 256 * 1024 * 1024;
-    pub const DEFAULT_CHECKPOINT_PERIOD: Duration = Duration::from_secs(1);
+    pub const DEFAULT_CHECKPOINT_PERIOD: Duration = Duration::from_secs(10);
 
-    pub const DEFAULT_GC_HORIZON: u64 = 64 * 1024 * 1024;
-    pub const DEFAULT_GC_PERIOD: Duration = Duration::from_secs(100);
+    pub const DEFAULT_UPLOAD_DISTANCE: u64 = 1024 * 1024 * 1024;
+    pub const DEFAULT_UPLOAD_PERIOD: Duration = Duration::from_secs(2500);
+
+    pub const DEFAULT_RECONSTRUCT_THRESHOLD: u64 = 0;
+
+    pub const DEFAULT_GC_HORIZON: u64 = 1024;
+    pub const DEFAULT_GC_PERIOD: Duration = Duration::from_secs(10);
 
     pub const DEFAULT_SUPERUSER: &str = "zenith_admin";
     pub const DEFAULT_RELISH_STORAGE_MAX_CONCURRENT_SYNC_LIMITS: usize = 100;
@@ -64,6 +69,9 @@ pub struct PageServerConf {
     // page server crashes.
     pub checkpoint_distance: u64,
     pub checkpoint_period: Duration,
+    pub upload_period: Duration,
+    pub upload_distance: u64,
+    pub reconstruct_threshold: u64,
 
     pub gc_horizon: u64,
     pub gc_period: Duration,
@@ -149,6 +157,9 @@ impl PageServerConf {
             daemonize: false,
             checkpoint_distance: defaults::DEFAULT_CHECKPOINT_DISTANCE,
             checkpoint_period: Duration::from_secs(10),
+            upload_distance: defaults::DEFAULT_UPLOAD_DISTANCE,
+            upload_period: defaults::DEFAULT_UPLOAD_PERIOD,
+            reconstruct_threshold: defaults::DEFAULT_RECONSTRUCT_THRESHOLD,
             gc_horizon: defaults::DEFAULT_GC_HORIZON,
             gc_period: Duration::from_secs(10),
             listen_pg_addr: defaults::DEFAULT_PG_LISTEN_ADDR.to_string(),

pageserver/src/page_service.rs

@@ -693,67 +693,21 @@ impl postgres_backend::Handler for PageServerHandler {
             let result = repo.gc_iteration(Some(timelineid), gc_horizon, true)?;
 
             pgb.write_message_noflush(&BeMessage::RowDescription(&[
-                RowDescriptor::int8_col(b"layer_relfiles_total"),
-                RowDescriptor::int8_col(b"layer_relfiles_needed_by_cutoff"),
-                RowDescriptor::int8_col(b"layer_relfiles_needed_by_branches"),
-                RowDescriptor::int8_col(b"layer_relfiles_not_updated"),
-                RowDescriptor::int8_col(b"layer_relfiles_needed_as_tombstone"),
-                RowDescriptor::int8_col(b"layer_relfiles_removed"),
-                RowDescriptor::int8_col(b"layer_relfiles_dropped"),
-                RowDescriptor::int8_col(b"layer_nonrelfiles_total"),
-                RowDescriptor::int8_col(b"layer_nonrelfiles_needed_by_cutoff"),
-                RowDescriptor::int8_col(b"layer_nonrelfiles_needed_by_branches"),
-                RowDescriptor::int8_col(b"layer_nonrelfiles_not_updated"),
-                RowDescriptor::int8_col(b"layer_nonrelfiles_needed_as_tombstone"),
-                RowDescriptor::int8_col(b"layer_nonrelfiles_removed"),
-                RowDescriptor::int8_col(b"layer_nonrelfiles_dropped"),
+                RowDescriptor::int8_col(b"meta_total"),
+                RowDescriptor::int8_col(b"meta_removed"),
+                RowDescriptor::int8_col(b"meta_dropped"),
+                RowDescriptor::int8_col(b"pages_total"),
+                RowDescriptor::int8_col(b"pages_removed"),
+                RowDescriptor::int8_col(b"pages_dropped"),
                 RowDescriptor::int8_col(b"elapsed"),
             ]))?
             .write_message_noflush(&BeMessage::DataRow(&[
-                Some(result.ondisk_relfiles_total.to_string().as_bytes()),
-                Some(
-                    result
-                        .ondisk_relfiles_needed_by_cutoff
-                        .to_string()
-                        .as_bytes(),
-                ),
-                Some(
-                    result
-                        .ondisk_relfiles_needed_by_branches
-                        .to_string()
-                        .as_bytes(),
-                ),
-                Some(result.ondisk_relfiles_not_updated.to_string().as_bytes()),
-                Some(
-                    result
-                        .ondisk_relfiles_needed_as_tombstone
-                        .to_string()
-                        .as_bytes(),
-                ),
-                Some(result.ondisk_relfiles_removed.to_string().as_bytes()),
-                Some(result.ondisk_relfiles_dropped.to_string().as_bytes()),
-                Some(result.ondisk_nonrelfiles_total.to_string().as_bytes()),
-                Some(
-                    result
-                        .ondisk_nonrelfiles_needed_by_cutoff
-                        .to_string()
-                        .as_bytes(),
-                ),
-                Some(
-                    result
-                        .ondisk_nonrelfiles_needed_by_branches
-                        .to_string()
-                        .as_bytes(),
-                ),
-                Some(result.ondisk_nonrelfiles_not_updated.to_string().as_bytes()),
-                Some(
-                    result
-                        .ondisk_nonrelfiles_needed_as_tombstone
-                        .to_string()
-                        .as_bytes(),
-                ),
-                Some(result.ondisk_nonrelfiles_removed.to_string().as_bytes()),
-                Some(result.ondisk_nonrelfiles_dropped.to_string().as_bytes()),
+                Some(result.meta_total.to_string().as_bytes()),
+                Some(result.meta_removed.to_string().as_bytes()),
+                Some(result.meta_dropped.to_string().as_bytes()),
+                Some(result.pages_total.to_string().as_bytes()),
+                Some(result.pages_removed.to_string().as_bytes()),
+                Some(result.pages_dropped.to_string().as_bytes()),
                 Some(result.elapsed.as_millis().to_string().as_bytes()),
             ]))?
             .write_message(&BeMessage::CommandComplete(b"SELECT 1"))?;

pageserver/src/relish_storage/rust_s3.rs

@@ -115,9 +115,9 @@ impl RelishStorage for RustS3 {
             .delete_object(path.key())
             .await
             .with_context(|| format!("Failed to delete s3 object with key {}", path.key()))?;
-        if code != 200 {
+        if code != 204 {
             Err(anyhow::format_err!(
-                "Received non-200 exit code during deleting object with key '{}', code: {}",
+                "Received non-204 exit code during deleting object with key '{}', code: {}",
                 path.key(),
                 code
             ))

pageserver/src/repository.rs

@@ -44,45 +44,27 @@ pub trait Repository: Send + Sync {
 ///
 /// Result of performing GC
 ///
-#[derive(Default)]
+#[derive(Default, Debug)]
 pub struct GcResult {
-    pub ondisk_relfiles_total: u64,
-    pub ondisk_relfiles_needed_by_cutoff: u64,
-    pub ondisk_relfiles_needed_by_branches: u64,
-    pub ondisk_relfiles_not_updated: u64,
-    pub ondisk_relfiles_needed_as_tombstone: u64,
-    pub ondisk_relfiles_removed: u64, // # of layer files removed because they have been made obsolete by newer ondisk files.
-    pub ondisk_relfiles_dropped: u64, // # of layer files removed because the relation was dropped
+    pub meta_removed: u64, // removed versions beyond PITR interval for which new page image exists
+    pub meta_dropped: u64, // removed versions beyond PITR interval of dropped relations
+    pub meta_total: u64,   // total number of metaobject version histories
 
-    pub ondisk_nonrelfiles_total: u64,
-    pub ondisk_nonrelfiles_needed_by_cutoff: u64,
-    pub ondisk_nonrelfiles_needed_by_branches: u64,
-    pub ondisk_nonrelfiles_not_updated: u64,
-    pub ondisk_nonrelfiles_needed_as_tombstone: u64,
-    pub ondisk_nonrelfiles_removed: u64, // # of layer files removed because they have been made obsolete by newer ondisk files.
-    pub ondisk_nonrelfiles_dropped: u64, // # of layer files removed because the relation was dropped
+    pub pages_removed: u64, // removed versions beyond PITR interval for which new page image exists
+    pub pages_dropped: u64, // removed versions beyond PITR interval of dropped relations
+    pub pages_total: u64,   // total number of page vaersion histories
 
     pub elapsed: Duration,
 }
 
 impl AddAssign for GcResult {
     fn add_assign(&mut self, other: Self) {
-        self.ondisk_relfiles_total += other.ondisk_relfiles_total;
-        self.ondisk_relfiles_needed_by_cutoff += other.ondisk_relfiles_needed_by_cutoff;
-        self.ondisk_relfiles_needed_by_branches += other.ondisk_relfiles_needed_by_branches;
-        self.ondisk_relfiles_not_updated += other.ondisk_relfiles_not_updated;
-        self.ondisk_relfiles_needed_as_tombstone += other.ondisk_relfiles_needed_as_tombstone;
-        self.ondisk_relfiles_removed += other.ondisk_relfiles_removed;
-        self.ondisk_relfiles_dropped += other.ondisk_relfiles_dropped;
-
-        self.ondisk_nonrelfiles_total += other.ondisk_nonrelfiles_total;
-        self.ondisk_nonrelfiles_needed_by_cutoff += other.ondisk_nonrelfiles_needed_by_cutoff;
-        self.ondisk_nonrelfiles_needed_by_branches += other.ondisk_nonrelfiles_needed_by_branches;
-        self.ondisk_nonrelfiles_not_updated += other.ondisk_nonrelfiles_not_updated;
-        self.ondisk_nonrelfiles_needed_as_tombstone += other.ondisk_nonrelfiles_needed_as_tombstone;
-        self.ondisk_nonrelfiles_removed += other.ondisk_nonrelfiles_removed;
-        self.ondisk_nonrelfiles_dropped += other.ondisk_nonrelfiles_dropped;
-
+        self.meta_total += other.meta_total;
+        self.meta_removed += other.meta_removed;
+        self.meta_dropped += other.meta_dropped;
+        self.pages_total += other.pages_total;
+        self.pages_removed += other.pages_removed;
+        self.pages_dropped += other.pages_dropped;
         self.elapsed += other.elapsed;
     }
 }
@@ -111,14 +93,21 @@ pub trait Timeline: Send + Sync {
 
     /// Get a list of all existing relations
     /// Pass RelTag to get relation objects or None to get nonrels.
-    fn list_relishes(&self, tag: Option<RelTag>, lsn: Lsn) -> Result<HashSet<RelishTag>>;
-
     /// Get a list of all existing relations in given tablespace and database.
     fn list_rels(&self, spcnode: u32, dbnode: u32, lsn: Lsn) -> Result<HashSet<RelishTag>>;
 
     /// Get a list of all existing non-relational objects
     fn list_nonrels(&self, lsn: Lsn) -> Result<HashSet<RelishTag>>;
 
+    ///
+    /// Export data as delats and image layers between 'start_lsn' to 'end_lsn'. The
+    /// start is inclusive, and end is exclusive.
+    ///
+    fn export_timeline(&self, start_lsn: Lsn, end_lsn: Lsn) -> Result<()>;
+
+    /// Get the LSN where this branch was created
+    fn get_ancestor_lsn(&self) -> Lsn;
+
     //------------------------------------------------------------------------------
     // Public PUT functions, to update the repository with new page versions.
     //
@@ -178,6 +167,16 @@ pub trait TimelineWriter: Deref<Target = dyn Timeline> {
     /// 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);
+
+    ///
+    /// Complete all delayed commits and advance disk_consistent_lsn
+    ///
+    fn checkpoint(&self) -> Result<()>;
+
+    ///
+    /// Import data from layer files
+    ///
+    fn import_timeline(&self, snapshot_lsn: Lsn) -> Result<()>;
 }
 
 #[derive(Debug, Clone, Serialize, Deserialize, PartialEq, Eq)]
@@ -210,6 +209,39 @@ impl WALRecord {
     }
 }
 
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub enum PageVersion {
+    /// an 8kb page image
+    Page(Bytes),
+    /// WAL record to get from previous page version to this one.
+    Wal(WALRecord),
+}
+
+///
+/// Data needed to reconstruct a page version
+///
+/// 'page_img' is the old base image of the page to start the WAL replay with.
+/// It can be None, if the first WAL record initializes the page (will_init)
+/// 'records' contains the records to apply over the base image.
+///
+pub struct PageReconstructData {
+    pub records: Vec<(Lsn, WALRecord)>,
+    pub page_img: Option<Bytes>,
+}
+
+/// Return value from Layer::get_page_reconstruct_data
+pub enum PageReconstructResult {
+    /// Got all the data needed to reconstruct the requested page
+    Complete,
+    /// This layer didn't contain all the required data, the caller should look up
+    /// the predecessor layer at the returned LSN and collect more data from there.
+    Continue(Lsn),
+    /// This layer didn't contain data needed to reconstruct the page version at
+    /// the returned LSN. This is usually considered an error, but might be OK
+    /// in some circumstances.
+    Missing(Lsn),
+}
+
 ///
 /// Tests that should work the same with any Repository/Timeline implementation.
 ///
@@ -217,7 +249,7 @@ impl WALRecord {
 #[cfg(test)]
 mod tests {
     use super::*;
-    use crate::layered_repository::{LayeredRepository, METADATA_FILE_NAME};
+    use crate::buffered_repository::{BufferedRepository, METADATA_FILE_NAME};
     use crate::walredo::{WalRedoError, WalRedoManager};
     use crate::PageServerConf;
     use hex_literal::hex;
@@ -293,7 +325,7 @@ mod tests {
         fn load(&self) -> Box<dyn Repository> {
             let walredo_mgr = Arc::new(TestRedoManager);
 
-            Box::new(LayeredRepository::new(
+            Box::new(BufferedRepository::new(
                 self.conf,
                 walredo_mgr,
                 self.tenant_id,

pageserver/src/restore_local_repo.rs

@@ -7,11 +7,11 @@ use postgres_ffi::nonrelfile_utils::slru_may_delete_clogsegment;
 use std::cmp::min;
 use std::fs;
 use std::fs::File;
-use std::io::Read;
-use std::path::Path;
+use std::io::{Read, Seek, SeekFrom};
+use std::path::{Path, PathBuf};
 
-use anyhow::{bail, Result};
-use bytes::{Buf, Bytes};
+use anyhow::{anyhow, bail, Result};
+use bytes::{Buf, Bytes, BytesMut};
 use tracing::*;
 
 use crate::relish::*;
@@ -37,6 +37,8 @@ pub fn import_timeline_from_postgres_datadir(
     writer: &dyn TimelineWriter,
     lsn: Lsn,
 ) -> Result<()> {
+    let mut pg_control: Option<ControlFileData> = None;
+
     // Scan 'global'
     for direntry in fs::read_dir(path.join("global"))? {
         let direntry = direntry?;
@@ -44,7 +46,7 @@ pub fn import_timeline_from_postgres_datadir(
             None => continue,
 
             Some("pg_control") => {
-                import_control_file(writer, lsn, &direntry.path())?;
+                pg_control = Some(import_control_file(writer, lsn, &direntry.path())?);
             }
             Some("pg_filenode.map") => import_nonrel_file(
                 writer,
@@ -124,9 +126,21 @@ pub fn import_timeline_from_postgres_datadir(
         import_nonrel_file(writer, lsn, RelishTag::TwoPhase { xid }, &entry.path())?;
     }
     // TODO: Scan pg_tblspc
-
     writer.advance_last_record_lsn(lsn);
 
+    // Import WAL. This is needed even when starting from a shutdown checkpoint, because
+    // this reads the checkpoint record itself, advancing the tip of the timeline to
+    // *after* the checkpoint record. And crucially, it initializes the 'prev_lsn'
+    let pg_control = pg_control.ok_or_else(|| anyhow!("pg_control file not found"))?;
+    import_wal(
+        &path.join("pg_wal"),
+        writer,
+        Lsn(pg_control.checkPointCopy.redo),
+        lsn,
+        &mut pg_control.checkPointCopy.clone(),
+    )?;
+    writer.checkpoint()?;
+
     Ok(())
 }
 
@@ -212,7 +226,11 @@ 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 TimelineWriter,
+    lsn: Lsn,
+    path: &Path,
+) -> Result<ControlFileData> {
     let mut file = File::open(path)?;
     let mut buffer = Vec::new();
     // read the whole file
@@ -233,7 +251,7 @@ fn import_control_file(timeline: &dyn TimelineWriter, lsn: Lsn, path: &Path) ->
     let checkpoint_bytes = pg_control.checkPointCopy.encode();
     timeline.put_page_image(RelishTag::Checkpoint, 0, lsn, checkpoint_bytes)?;
 
-    Ok(())
+    Ok(pg_control)
 }
 
 ///
@@ -285,6 +303,104 @@ fn import_slru_file(
     Ok(())
 }
 
+/// Scan PostgreSQL WAL files in given directory and load all records between
+/// 'startpoint' and 'endpoint' into the repository.
+fn import_wal(
+    walpath: &Path,
+    timeline: &dyn TimelineWriter,
+    startpoint: Lsn,
+    endpoint: Lsn,
+    checkpoint: &mut CheckPoint,
+) -> Result<()> {
+    let mut waldecoder = WalStreamDecoder::new(startpoint);
+
+    let mut segno = startpoint.segment_number(pg_constants::WAL_SEGMENT_SIZE);
+    let mut offset = startpoint.segment_offset(pg_constants::WAL_SEGMENT_SIZE);
+    let mut last_lsn = startpoint;
+
+    while last_lsn <= endpoint {
+        // FIXME: assume postgresql tli 1 for now
+        let filename = XLogFileName(1, segno, pg_constants::WAL_SEGMENT_SIZE);
+        let mut buf = Vec::new();
+
+        // Read local file
+        let mut path = walpath.join(&filename);
+
+        // It could be as .partial
+        if !PathBuf::from(&path).exists() {
+            path = walpath.join(filename + ".partial");
+        }
+
+        // Slurp the WAL file
+        let mut file = File::open(&path)?;
+
+        if offset > 0 {
+            file.seek(SeekFrom::Start(offset as u64))?;
+        }
+
+        let nread = file.read_to_end(&mut buf)?;
+        if nread != pg_constants::WAL_SEGMENT_SIZE - offset as usize {
+            // Maybe allow this for .partial files?
+            error!("read only {} bytes from WAL file", nread);
+        }
+
+        waldecoder.feed_bytes(&buf);
+
+        let mut nrecords = 0;
+        while last_lsn <= endpoint {
+            if let Some((lsn, recdata)) = waldecoder.poll_decode()? {
+                let mut checkpoint_modified = false;
+
+                let decoded = decode_wal_record(recdata.clone());
+                save_decoded_record(
+                    checkpoint,
+                    &mut checkpoint_modified,
+                    timeline,
+                    &decoded,
+                    recdata,
+                    lsn,
+                )?;
+                last_lsn = lsn;
+
+                if checkpoint_modified {
+                    let checkpoint_bytes = checkpoint.encode();
+                    timeline.put_page_image(
+                        RelishTag::Checkpoint,
+                        0,
+                        last_lsn,
+                        checkpoint_bytes,
+                    )?;
+                }
+
+                // 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
+                timeline.advance_last_record_lsn(last_lsn);
+                nrecords += 1;
+
+                trace!("imported record at {} (end {})", lsn, endpoint);
+            }
+        }
+
+        debug!("imported {} records up to {}", nrecords, last_lsn);
+
+        segno += 1;
+        offset = 0;
+    }
+
+    if last_lsn != startpoint {
+        debug!(
+            "reached end of WAL at {}, updating checkpoint info",
+            last_lsn
+        );
+
+        timeline.advance_last_record_lsn(last_lsn);
+    } else {
+        info!("no WAL to import at {}", last_lsn);
+    }
+
+    Ok(())
+}
+
 ///
 /// Helper function to parse a WAL record and call the Timeline's PUT functions for all the
 /// relations/pages that the record affects.
@@ -300,7 +416,6 @@ pub fn save_decoded_record(
     if checkpoint.update_next_xid(decoded.xl_xid) {
         *checkpoint_modified = true;
     }
-
     // Iterate through all the blocks that the record modifies, and
     // "put" a separate copy of the record for each block.
     for blk in decoded.blocks.iter() {
@@ -310,14 +425,38 @@ pub fn save_decoded_record(
             relnode: blk.rnode_relnode,
             forknum: blk.forknum as u8,
         });
+        if blk.apply_image
+            && blk.has_image
+            && decoded.xl_rmid == pg_constants::RM_XLOG_ID
+            && (decoded.xl_info == pg_constants::XLOG_FPI
+                || decoded.xl_info == pg_constants::XLOG_FPI_FOR_HINT)
+        {
+            // Extract page image from FPI record
+            let img_len = blk.bimg_len as usize;
+            let img_offs = blk.bimg_offset as usize;
+            let mut image = BytesMut::with_capacity(pg_constants::BLCKSZ as usize);
+            image.extend_from_slice(&recdata[img_offs..img_offs + img_len]);
 
-        let rec = WALRecord {
-            will_init: blk.will_init || blk.apply_image,
-            rec: recdata.clone(),
-            main_data_offset: decoded.main_data_offset as u32,
-        };
+            // Compression of WAL is not yet supported
+            assert!((blk.bimg_info & pg_constants::BKPIMAGE_IS_COMPRESSED) == 0);
 
-        timeline.put_wal_record(lsn, tag, blk.blkno, rec)?;
+            if blk.hole_length != 0 {
+                let tail = image.split_off(blk.hole_offset as usize);
+                image.resize(image.len() + blk.hole_length as usize, 0u8);
+                image.unsplit(tail);
+            }
+            image[0..4].copy_from_slice(&((lsn.0 >> 32) as u32).to_le_bytes());
+            image[4..8].copy_from_slice(&(lsn.0 as u32).to_le_bytes());
+            assert_eq!(image.len(), pg_constants::BLCKSZ as usize);
+            timeline.put_page_image(tag, blk.blkno, lsn, image.freeze())?;
+        } else {
+            let rec = WALRecord {
+                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)?;
+        }
     }
 
     let mut buf = decoded.record.clone();

pageserver/src/tenant_mgr.rs

@@ -2,15 +2,15 @@
 //! page server.
 
 use crate::branches;
-use crate::layered_repository::LayeredRepository;
+use crate::buffered_repository::BufferedRepository;
 use crate::repository::{Repository, Timeline};
 use crate::walredo::PostgresRedoManager;
 use crate::PageServerConf;
 use anyhow::{anyhow, bail, Context, Result};
 use lazy_static::lazy_static;
 use log::{debug, info};
-use std::collections::hash_map::Entry;
 use std::collections::HashMap;
+use std::fmt;
 use std::fs;
 use std::str::FromStr;
 use std::sync::atomic::{AtomicBool, Ordering};
@@ -19,15 +19,50 @@ use std::thread::JoinHandle;
 use zenith_utils::zid::{ZTenantId, ZTimelineId};
 
 lazy_static! {
-    static ref REPOSITORY: Mutex<HashMap<ZTenantId, Arc<dyn Repository>>> =
-        Mutex::new(HashMap::new());
+    static ref TENANTS: Mutex<HashMap<ZTenantId, Tenant>> = Mutex::new(HashMap::new());
 }
 
-fn access_repository() -> MutexGuard<'static, HashMap<ZTenantId, Arc<dyn Repository>>> {
-    REPOSITORY.lock().unwrap()
+struct Tenant {
+    state: TenantState,
+    repo: Option<Arc<dyn Repository>>,
 }
+
+#[derive(Debug)]
+enum TenantState {
+    // This tenant only exists in cloud storage. It cannot be accessed.
+    CloudOnly,
+    // This tenant exists in cloud storage, and we are currently downloading it to local disk.
+    // It cannot be accessed yet, not until it's been fully downloaded to local disk.
+    Downloading,
+    // All data for this tenant is complete on local disk, but we haven't loaded the Repository,
+    // Timeline and Layer structs into memory yet, so it cannot be accessed yet.
+    //Ready,
+    // This tenant exists on local disk, and the layer map has been loaded into memory.
+    // The local disk might have some newer files that don't exist in cloud storage yet.
+    Active,
+    // This tenant exists on local disk, and the layer map has been loaded into memory.
+    // The local disk might have some newer files that don't exist in cloud storage yet.
+    // The tenant cannot be accessed anymore for any reason, but graceful shutdown.
+    //Stopping,
+}
+
+impl fmt::Display for TenantState {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        match self {
+            TenantState::CloudOnly => f.write_str("CloudOnly"),
+            TenantState::Downloading => f.write_str("Downloading"),
+            TenantState::Active => f.write_str("Active"),
+        }
+    }
+}
+
+fn access_tenants() -> MutexGuard<'static, HashMap<ZTenantId, Tenant>> {
+    TENANTS.lock().unwrap()
+}
+
 struct TenantHandleEntry {
     checkpointer_handle: Option<JoinHandle<()>>,
+    uploader_handle: Option<JoinHandle<()>>,
     gc_handle: Option<JoinHandle<()>>,
 }
 
@@ -41,40 +76,53 @@ lazy_static! {
 static SHUTDOWN_REQUESTED: AtomicBool = AtomicBool::new(false);
 
 pub fn init(conf: &'static PageServerConf) {
-    let mut m = access_repository();
     for dir_entry in fs::read_dir(conf.tenants_path()).unwrap() {
         let tenantid =
             ZTenantId::from_str(dir_entry.unwrap().file_name().to_str().unwrap()).unwrap();
-        let repo = init_repo(conf, tenantid);
+
+        {
+            let mut m = access_tenants();
+            let tenant = Tenant {
+                state: TenantState::CloudOnly,
+                repo: None,
+            };
+            m.insert(tenantid, tenant);
+        }
+
+        init_repo(conf, tenantid);
         info!("initialized storage for tenant: {}", &tenantid);
-        m.insert(tenantid, repo);
     }
 }
 
-fn init_repo(conf: &'static PageServerConf, tenant_id: ZTenantId) -> Arc<LayeredRepository> {
+fn init_repo(conf: &'static PageServerConf, tenant_id: ZTenantId) {
     // Set up a WAL redo manager, for applying WAL records.
     let walredo_mgr = PostgresRedoManager::new(conf, tenant_id);
 
     // Set up an object repository, for actual data storage.
-    let repo = Arc::new(LayeredRepository::new(
+    let repo = Arc::new(BufferedRepository::new(
         conf,
         Arc::new(walredo_mgr),
         tenant_id,
         true,
     ));
 
-    let checkpointer_handle = LayeredRepository::launch_checkpointer_thread(conf, repo.clone());
-    let gc_handle = LayeredRepository::launch_gc_thread(conf, repo.clone());
+    let checkpointer_handle = BufferedRepository::launch_checkpointer_thread(conf, repo.clone());
+    let gc_handle = BufferedRepository::launch_gc_thread(conf, repo.clone());
+    let uploader_handle = BufferedRepository::launch_upload_thread(conf, repo.clone());
 
     let mut handles = TENANT_HANDLES.lock().unwrap();
     let h = TenantHandleEntry {
         checkpointer_handle: Some(checkpointer_handle),
         gc_handle: Some(gc_handle),
+        uploader_handle: Some(uploader_handle),
     };
 
     handles.insert(tenant_id, h);
 
-    repo
+    let mut m = access_tenants();
+    let tenant = m.get_mut(&tenant_id).unwrap();
+    tenant.repo = Some(repo);
+    tenant.state = TenantState::Active;
 }
 
 // TODO kb Currently unused function, will later be used when the relish storage downloads a new layer.
@@ -89,15 +137,23 @@ pub fn register_relish_download(
         tenant_id,
         timeline_id
     );
-    match access_repository().entry(tenant_id) {
-        Entry::Occupied(o) => init_timeline(o.get().as_ref(), timeline_id),
-        Entry::Vacant(v) => {
-            log::info!("New repo initialized");
-            let new_repo = init_repo(conf, tenant_id);
-            init_timeline(new_repo.as_ref(), timeline_id);
-            v.insert(new_repo);
+
+    {
+        let mut m = access_tenants();
+        let mut tenant = m.get_mut(&tenant_id).unwrap();
+        tenant.state = TenantState::Downloading;
+        match &tenant.repo {
+            Some(repo) => init_timeline(repo.as_ref(), timeline_id),
+            None => {
+                log::info!("Initialize new repo");
+            }
         }
     }
+
+    // init repo updates Tenant state
+    init_repo(conf, tenant_id);
+    let new_repo = get_repository_for_tenant(tenant_id).unwrap();
+    init_timeline(new_repo.as_ref(), timeline_id);
 }
 
 fn init_timeline(repo: &dyn Repository, timeline_id: ZTimelineId) {
@@ -117,6 +173,8 @@ pub fn stop_tenant_threads(tenantid: ZTenantId) {
     if let Some(h) = handles.get_mut(&tenantid) {
         h.checkpointer_handle.take().map(JoinHandle::join);
         debug!("checkpointer for tenant {} has stopped", tenantid);
+        h.uploader_handle.take().map(JoinHandle::join);
+        debug!("uploader for tenant {} has stopped", tenantid);
         h.gc_handle.take().map(JoinHandle::join);
         debug!("gc for tenant {} has stopped", tenantid);
     }
@@ -125,8 +183,8 @@ pub fn stop_tenant_threads(tenantid: ZTenantId) {
 pub fn shutdown_all_tenants() -> Result<()> {
     SHUTDOWN_REQUESTED.swap(true, Ordering::Relaxed);
 
-    let tenants = list_tenants()?;
-    for tenantid in tenants {
+    let tenantids = list_tenantids()?;
+    for tenantid in tenantids {
         stop_tenant_threads(tenantid);
         let repo = get_repository_for_tenant(tenantid)?;
         debug!("shutdown tenant {}", tenantid);
@@ -140,25 +198,40 @@ pub fn create_repository_for_tenant(
     conf: &'static PageServerConf,
     tenantid: ZTenantId,
 ) -> Result<()> {
-    let mut m = access_repository();
-
-    // First check that the tenant doesn't exist already
-    if m.get(&tenantid).is_some() {
-        bail!("tenant {} already exists", tenantid);
+    {
+        let mut m = access_tenants();
+        // First check that the tenant doesn't exist already
+        if m.get(&tenantid).is_some() {
+            bail!("tenant {} already exists", tenantid);
+        }
+        let tenant = Tenant {
+            state: TenantState::CloudOnly,
+            repo: None,
+        };
+        m.insert(tenantid, tenant);
     }
+
     let wal_redo_manager = Arc::new(PostgresRedoManager::new(conf, tenantid));
     let repo = branches::create_repo(conf, tenantid, wal_redo_manager)?;
 
-    m.insert(tenantid, repo);
+    let mut m = access_tenants();
+    let tenant = m.get_mut(&tenantid).unwrap();
+    tenant.repo = Some(repo);
+    tenant.state = TenantState::Active;
 
     Ok(())
 }
 
 pub fn get_repository_for_tenant(tenantid: ZTenantId) -> Result<Arc<dyn Repository>> {
-    access_repository()
+    let m = access_tenants();
+    let tenant = m
         .get(&tenantid)
-        .map(Arc::clone)
-        .ok_or_else(|| anyhow!("repository not found for tenant name {}", tenantid))
+        .ok_or_else(|| anyhow!("Tenant not found for tenant {}", tenantid));
+
+    match &tenant.unwrap().repo {
+        Some(repo) => Ok(Arc::clone(repo)),
+        None => anyhow::bail!("Repository for tenant {} is not yet valid", tenantid),
+    }
 }
 
 pub fn get_timeline_for_tenant(
@@ -170,12 +243,11 @@ pub fn get_timeline_for_tenant(
         .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;
+fn list_tenantids() -> Result<Vec<ZTenantId>> {
+    let m = access_tenants();
+    m.iter()
+        .map(|v| {
+            let (tenantid, _) = v;
             Ok(*tenantid)
         })
         .collect()

pageserver/src/toast_store.rs

@@ -0,0 +1,258 @@
+use anyhow::{anyhow, Result};
+use lz4_flex;
+use std::convert::TryInto;
+use std::ops::{Bound, RangeBounds};
+use std::path::Path;
+
+use yakv::storage::{Key, Storage, StorageConfig, StorageIterator, Transaction, Value};
+
+const TOAST_SEGMENT_SIZE: usize = 2 * 1024;
+const CACHE_SIZE: usize = 32 * 1024; // 256Mb
+                                     //const CACHE_SIZE: usize = 128 * 1024; // 1Gb
+
+///
+/// Toast storage consistof two KV databases: one for storing main index
+/// and second for storing sliced BLOB (values larger than 2kb).
+/// BLOBs and main data are stored in different databases to improve
+/// data locality and reduce key size for TOAST segments.
+///
+pub struct ToastStore {
+    db: Storage,         // key-value database
+}
+
+pub struct ToastIterator<'a> {
+    iter: StorageIterator<'a>,
+}
+
+#[derive(Clone, Copy)]
+pub struct PageData {
+    data: [u8; 8192],
+}
+
+impl PageData {
+    pub fn find_first_zero_bit(&self, offs: usize) -> usize {
+        let bytes = self.data;
+        for i in offs..8192 {
+            if bytes[i] != 0xFFu8 {
+                return i * 8 + bytes[i].trailing_ones() as usize;
+            }
+        }
+        usize::MAX
+    }
+}
+
+impl<'a> Iterator for ToastIterator<'a> {
+    type Item = Result<(Key, Value)>;
+    fn next(&mut self) -> Option<Self::Item> {
+        let mut toast: Option<Vec<u8>> = None;
+        let mut next_segno = 0u16;
+        for elem in &mut self.iter {
+            if let Ok((key, value)) = elem {
+                let key_len = key.len();
+                let n_segments =
+                    u16::from_be_bytes(key[key_len - 4..key_len - 2].try_into().unwrap());
+                let segno = u16::from_be_bytes(key[key_len - 2..].try_into().unwrap());
+                let key = key[..key_len - 4].to_vec();
+                if n_segments != 0 {
+                    // TOAST
+                    assert_eq!(segno, next_segno);
+                    if next_segno == 0 {
+                        toast = Some(Vec::with_capacity(n_segments as usize * TOAST_SEGMENT_SIZE))
+                    }
+                    toast.as_mut().unwrap().extend_from_slice(&value);
+                    next_segno = segno + 1;
+                    if next_segno == n_segments {
+                        let res = lz4_flex::decompress_size_prepended(&toast.unwrap());
+                        return Some(if let Ok(decompressed_data) = res {
+                            Ok((key, decompressed_data))
+                        } else {
+                            Err(anyhow!(res.unwrap_err()))
+                        });
+                    }
+                } else {
+                    return Some(Ok((key, value)));
+                }
+            } else {
+                return Some(elem);
+            }
+        }
+        assert_eq!(next_segno, 0);
+        None
+    }
+}
+
+impl<'a> DoubleEndedIterator for ToastIterator<'a> {
+    fn next_back(&mut self) -> Option<Self::Item> {
+        let mut toast: Option<Vec<u8>> = None;
+        let mut next_segno = 0u16;
+        while let Some(elem) = self.iter.next_back() {
+            if let Ok((key, value)) = elem {
+                assert!(!value.is_empty());
+                let key_len = key.len();
+                let n_segments =
+                    u16::from_be_bytes(key[key_len - 4..key_len - 2].try_into().unwrap());
+                let segno = u16::from_be_bytes(key[key_len - 2..].try_into().unwrap());
+                let key = key[..key_len - 4].to_vec();
+                if n_segments != 0 {
+                    // TOAST
+                    assert!(segno + 1 == next_segno || next_segno == 0);
+                    if next_segno == 0 {
+                        let len = (n_segments - 1) as usize * TOAST_SEGMENT_SIZE + value.len();
+                        let mut vec = vec![0u8; len];
+                        vec[len - value.len()..].copy_from_slice(&value);
+                        toast = Some(vec);
+                    } else {
+                        toast.as_mut().unwrap()[segno as usize * TOAST_SEGMENT_SIZE
+                            ..(segno + 1) as usize * TOAST_SEGMENT_SIZE]
+                            .copy_from_slice(&value);
+                    }
+                    next_segno = segno;
+                    if next_segno == 0 {
+                        let toast = toast.unwrap();
+                        assert!(!toast.is_empty());
+                        let res = lz4_flex::decompress_size_prepended(&toast);
+                        return Some(if let Ok(decompressed_data) = res {
+                            Ok((key, decompressed_data))
+                        } else {
+                            Err(anyhow!(res.unwrap_err()))
+                        });
+                    }
+                } else {
+                    return Some(Ok((key, value)));
+                }
+            } else {
+                return Some(elem);
+            }
+        }
+        assert_eq!(next_segno, 0);
+        None
+    }
+}
+
+//
+// FIXME-KK: not using WAL now. Implement asynchronous or delayed commit.
+//
+impl ToastStore {
+    pub fn new(path: &Path) -> Result<ToastStore> {
+        Ok(ToastStore {
+            db: Storage::open(
+                &path.join("pageserver.db"),
+                StorageConfig {
+                    cache_size: CACHE_SIZE,
+                    nosync: false,
+					mursiw: true,
+                },
+            )?,
+        })
+    }
+
+    pub fn put(&self, key: Key, value: Value) -> Result<()> {
+        let mut tx = self.db.start_transaction();
+		self.tx_remove(&mut tx, &key)?;
+        let value_len = value.len();
+        let mut key = key;
+        if value_len >= TOAST_SEGMENT_SIZE {
+            let compressed_data = lz4_flex::compress_prepend_size(&value);
+            let compressed_data_len = compressed_data.len();
+            let mut offs: usize = 0;
+            let mut segno = 0u16;
+            let n_segments =
+                ((compressed_data_len + TOAST_SEGMENT_SIZE - 1) / TOAST_SEGMENT_SIZE) as u16;
+            assert!(n_segments != 0);
+            key.extend_from_slice(&n_segments.to_be_bytes());
+            key.extend_from_slice(&[0u8; 2]);
+            let key_len = key.len();
+            while offs + TOAST_SEGMENT_SIZE < compressed_data_len {
+                key[key_len - 2..].copy_from_slice(&segno.to_be_bytes());
+                tx.put(
+                    &key,
+                    &compressed_data[offs..offs + TOAST_SEGMENT_SIZE].to_vec(),
+                )?;
+                offs += TOAST_SEGMENT_SIZE;
+                segno += 1;
+            }
+            key[key_len - 2..].copy_from_slice(&segno.to_be_bytes());
+            tx.put(&key, &compressed_data[offs..].to_vec())?;
+        } else {
+            key.extend_from_slice(&[0u8; 4]);
+            tx.put(&key, &value)?;
+        }
+        tx.delay()?;
+        Ok(())
+    }
+
+    pub fn commit(&mut self) -> Result<()> {
+        let tx = self.db.start_transaction();
+        tx.commit()?;
+        Ok(())
+    }
+
+    pub fn iter(&self) -> ToastIterator<'_> {
+        self.range(..)
+    }
+
+    pub fn range<R: RangeBounds<Key>>(&self, range: R) -> ToastIterator<'_> {
+        let from = match range.start_bound() {
+            Bound::Included(key) => {
+                let mut key = key.clone();
+                key.extend_from_slice(&[0u8; 4]);
+                Bound::Included(key)
+            }
+            Bound::Excluded(key) => {
+                let mut key = key.clone();
+                key.extend_from_slice(&[0u8; 4]);
+                Bound::Excluded(key)
+            }
+            _ => Bound::Unbounded,
+        };
+        let till = match range.end_bound() {
+            Bound::Included(key) => {
+                let mut key = key.clone();
+                key.extend_from_slice(&[0xFFu8; 4]);
+                Bound::Included(key)
+            }
+            Bound::Excluded(key) => {
+                let mut key = key.clone();
+                key.extend_from_slice(&[0xFFu8; 4]);
+                Bound::Excluded(key)
+            }
+            _ => Bound::Unbounded,
+        };
+        ToastIterator {
+            iter: self.db.range((from, till)),
+        }
+    }
+
+    pub fn remove(&self, key: Key) -> Result<()> {
+        let mut tx = self.db.start_transaction();
+        self.tx_remove(&mut tx, &key)?;
+        tx.delay()
+    }
+
+    pub fn tx_remove(&self, tx: &mut Transaction, key: &[u8]) -> Result<()> {
+        let mut min_key = key.to_vec();
+        let mut max_key = key.to_vec();
+        min_key.extend_from_slice(&[0u8; 4]);
+        max_key.extend_from_slice(&[0xFFu8; 4]);
+        let mut iter = tx.range(&min_key..&max_key);
+        if let Some(entry) = iter.next() {
+            let mut key = entry?.0;
+            let key_len = key.len();
+            let n_segments = u16::from_be_bytes(key[key_len - 4..key_len - 2].try_into().unwrap());
+            if n_segments != 0 {
+                // TOAST
+                for i in 0..n_segments {
+                    key[key_len - 2..].copy_from_slice(&i.to_be_bytes());
+                    tx.remove(&key)?;
+                }
+            } else {
+                tx.remove(&key)?;
+            }
+        }
+        Ok(())
+    }
+
+    pub fn size(&self) -> u64 {
+        self.db.get_database_info().db_used
+    }
+}

pageserver/src/waldecoder.rs

@@ -229,17 +229,18 @@ pub struct DecodedBkpBlock {
     pub blkno: u32,
 
     /* copy of the fork_flags field from the XLogRecordBlockHeader */
-    flags: u8,
+    pub flags: u8,
 
     /* Information on full-page image, if any */
-    has_image: bool,       /* has image, even for consistency checking */
+    pub has_image: bool,   /* has image, even for consistency checking */
     pub apply_image: bool, /* has image that should be restored */
     pub will_init: bool,   /* record doesn't need previous page version to apply */
     //char	   *bkp_image;
-    hole_offset: u16,
-    hole_length: u16,
-    bimg_len: u16,
-    bimg_info: u8,
+    pub hole_offset: u16,
+    pub hole_length: u16,
+    pub bimg_offset: u32,
+    pub bimg_len: u16,
+    pub bimg_info: u8,
 
     /* Buffer holding the rmgr-specific data associated with this block */
     has_data: bool,
@@ -859,8 +860,19 @@ pub fn decode_wal_record(record: Bytes) -> DecodedWALRecord {
     }
 
     // 3. Decode blocks.
+    let mut ptr = record.len() - buf.remaining();
+    for blk in blocks.iter_mut() {
+        if blk.has_image {
+            blk.bimg_offset = ptr as u32;
+            ptr += blk.bimg_len as usize;
+        }
+        if blk.has_data {
+            ptr += blk.data_len as usize;
+        }
+    }
     // We don't need them, so just skip blocks_total_len bytes
     buf.advance(blocks_total_len as usize);
+    assert_eq!(ptr, record.len() - buf.remaining());
 
     let main_data_offset = (xlogrec.xl_tot_len - main_data_len) as usize;
 

pageserver/src/walredo.rs

@@ -22,6 +22,7 @@ use byteorder::{ByteOrder, LittleEndian};
 use bytes::{Buf, BufMut, Bytes, BytesMut};
 use lazy_static::lazy_static;
 use log::*;
+use rand::Rng;
 use serde::Serialize;
 use std::fs;
 use std::fs::OpenOptions;
@@ -53,6 +54,8 @@ use postgres_ffi::nonrelfile_utils::transaction_id_set_status;
 use postgres_ffi::pg_constants;
 use postgres_ffi::XLogRecord;
 
+const WAL_REDO_WORKERS: usize = 1;
+
 ///
 /// `RelTag` + block number (`blknum`) gives us a unique id of the page in the cluster.
 ///
@@ -140,7 +143,7 @@ pub struct PostgresRedoManager {
     conf: &'static PageServerConf,
 
     runtime: tokio::runtime::Runtime,
-    process: Mutex<Option<PostgresRedoProcess>>,
+    workers: [Mutex<Option<PostgresRedoProcess>>; WAL_REDO_WORKERS],
 }
 
 #[derive(Debug)]
@@ -195,7 +198,9 @@ impl WalRedoManager for PostgresRedoManager {
 
         start_time = Instant::now();
         let result = {
-            let mut process_guard = self.process.lock().unwrap();
+            let mut process_guard = self.workers[rand::thread_rng().gen_range(0..WAL_REDO_WORKERS)]
+                .lock()
+                .unwrap();
             lock_time = Instant::now();
 
             // launch the WAL redo process on first use
@@ -237,7 +242,7 @@ impl PostgresRedoManager {
             runtime,
             tenantid,
             conf,
-            process: Mutex::new(None),
+            workers: [(); WAL_REDO_WORKERS].map(|_| Mutex::new(None)),
         }
     }
 

postgres_ffi/src/xlog_utils.rs

@@ -9,7 +9,6 @@
 
 use crate::pg_constants;
 use crate::CheckPoint;
-use crate::ControlFileData;
 use crate::FullTransactionId;
 use crate::XLogLongPageHeaderData;
 use crate::XLogPageHeaderData;
@@ -18,8 +17,8 @@ use crate::XLOG_PAGE_MAGIC;
 
 use anyhow::{bail, Result};
 use byteorder::{ByteOrder, LittleEndian};
+use bytes::BytesMut;
 use bytes::{Buf, Bytes};
-use bytes::{BufMut, BytesMut};
 use crc32c::*;
 use log::*;
 use std::cmp::max;
@@ -410,27 +409,25 @@ impl CheckPoint {
 }
 
 //
-// Generate new WAL segment with single XLOG_CHECKPOINT_SHUTDOWN record.
+// Generate new, empty WAL segment.
 // We need this segment to start compute node.
-// In order to minimize changes in Postgres core, we prefer to
-// provide WAL segment from which is can extract checkpoint record in standard way,
-// rather then implement some alternative mechanism.
 //
-pub fn generate_wal_segment(pg_control: &ControlFileData) -> Bytes {
+pub fn generate_wal_segment(segno: u64, system_id: u64) -> Bytes {
     let mut seg_buf = BytesMut::with_capacity(pg_constants::WAL_SEGMENT_SIZE as usize);
 
+    let pageaddr = XLogSegNoOffsetToRecPtr(segno, 0, pg_constants::WAL_SEGMENT_SIZE);
     let hdr = XLogLongPageHeaderData {
         std: {
             XLogPageHeaderData {
                 xlp_magic: XLOG_PAGE_MAGIC as u16,
                 xlp_info: pg_constants::XLP_LONG_HEADER,
                 xlp_tli: 1, // FIXME: always use Postgres timeline 1
-                xlp_pageaddr: pg_control.checkPoint - XLOG_SIZE_OF_XLOG_LONG_PHD as u64,
+                xlp_pageaddr: pageaddr,
                 xlp_rem_len: 0,
                 ..Default::default() // Put 0 in padding fields.
             }
         },
-        xlp_sysid: pg_control.system_identifier,
+        xlp_sysid: system_id,
         xlp_seg_size: pg_constants::WAL_SEGMENT_SIZE as u32,
         xlp_xlog_blcksz: XLOG_BLCKSZ as u32,
     };
@@ -438,36 +435,6 @@ pub fn generate_wal_segment(pg_control: &ControlFileData) -> Bytes {
     let hdr_bytes = hdr.encode();
     seg_buf.extend_from_slice(&hdr_bytes);
 
-    let rec_hdr = XLogRecord {
-        xl_tot_len: (XLOG_SIZE_OF_XLOG_RECORD
-            + SIZE_OF_XLOG_RECORD_DATA_HEADER_SHORT
-            + SIZEOF_CHECKPOINT) as u32,
-        xl_xid: 0, //0 is for InvalidTransactionId
-        xl_prev: 0,
-        xl_info: pg_constants::XLOG_CHECKPOINT_SHUTDOWN,
-        xl_rmid: pg_constants::RM_XLOG_ID,
-        xl_crc: 0,
-        ..Default::default() // Put 0 in padding fields.
-    };
-
-    let mut rec_shord_hdr_bytes = BytesMut::new();
-    rec_shord_hdr_bytes.put_u8(pg_constants::XLR_BLOCK_ID_DATA_SHORT);
-    rec_shord_hdr_bytes.put_u8(SIZEOF_CHECKPOINT as u8);
-
-    let rec_bytes = rec_hdr.encode();
-    let checkpoint_bytes = pg_control.checkPointCopy.encode();
-
-    //calculate record checksum
-    let mut crc = 0;
-    crc = crc32c_append(crc, &rec_shord_hdr_bytes[..]);
-    crc = crc32c_append(crc, &checkpoint_bytes[..]);
-    crc = crc32c_append(crc, &rec_bytes[0..XLOG_RECORD_CRC_OFFS]);
-
-    seg_buf.extend_from_slice(&rec_bytes[0..XLOG_RECORD_CRC_OFFS]);
-    seg_buf.put_u32_le(crc);
-    seg_buf.extend_from_slice(&rec_shord_hdr_bytes);
-    seg_buf.extend_from_slice(&checkpoint_bytes);
-
     //zero out the rest of the file
     seg_buf.resize(pg_constants::WAL_SEGMENT_SIZE, 0);
     seg_buf.freeze()

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")

test_runner/Pipfile

@@ -11,11 +11,14 @@ pyjwt = {extras = ["crypto"], version = "*"}
 requests = "*"
 pytest-xdist = "*"
 asyncpg = "*"
+cached-property = "*"
 
 [dev-packages]
-yapf = "*"
 flake8 = "*"
 mypy = "*"
+# Behavior may change slightly between versions. These are run continuously,
+# so we pin exact versions to avoid suprising breaks. Update if comfortable.
+yapf = "==0.31.0"
 
 [requires]
 # we need at least 3.6, but pipenv doesn't allow to say this directly

test_runner/Pipfile.lock

@@ -1,7 +1,7 @@
 {
     "_meta": {
         "hash": {
-            "sha256": "3cdc048691824d0b93912b6b78a0aa01dc98f278212c1badb0cc2edbd2103c3a"
+            "sha256": "3645ae8d2dcf55bd2a54963c44cfeedf577f3b289d1077365214a80a7f36e643"
         },
         "pipfile-spec": 6,
         "requires": {
@@ -43,94 +43,108 @@
             "markers": "python_version >= '2.7' and python_version not in '3.0, 3.1, 3.2, 3.3, 3.4'",
             "version": "==21.2.0"
         },
+        "cached-property": {
+            "hashes": [
+                "sha256:9fa5755838eecbb2d234c3aa390bd80fbd3ac6b6869109bfc1b499f7bd89a130",
+                "sha256:df4f613cf7ad9a588cc381aaf4a512d26265ecebd5eb9e1ba12f1319eb85a6a0"
+            ],
+            "index": "pypi",
+            "version": "==1.5.2"
+        },
         "certifi": {
             "hashes": [
-                "sha256:2bbf76fd432960138b3ef6dda3dde0544f27cbf8546c458e60baf371917ba9ee",
-                "sha256:50b1e4f8446b06f41be7dd6338db18e0990601dce795c2b1686458aa7e8fa7d8"
+                "sha256:78884e7c1d4b00ce3cea67b44566851c4343c120abd683433ce934a68ea58872",
+                "sha256:d62a0163eb4c2344ac042ab2bdf75399a71a2d8c7d47eac2e2ee91b9d6339569"
             ],
-            "version": "==2021.5.30"
+            "version": "==2021.10.8"
         },
         "cffi": {
             "hashes": [
-                "sha256:06c54a68935738d206570b20da5ef2b6b6d92b38ef3ec45c5422c0ebaf338d4d",
-                "sha256:0c0591bee64e438883b0c92a7bed78f6290d40bf02e54c5bf0978eaf36061771",
-                "sha256:19ca0dbdeda3b2615421d54bef8985f72af6e0c47082a8d26122adac81a95872",
-                "sha256:22b9c3c320171c108e903d61a3723b51e37aaa8c81255b5e7ce102775bd01e2c",
-                "sha256:26bb2549b72708c833f5abe62b756176022a7b9a7f689b571e74c8478ead51dc",
-                "sha256:33791e8a2dc2953f28b8d8d300dde42dd929ac28f974c4b4c6272cb2955cb762",
-                "sha256:3c8d896becff2fa653dc4438b54a5a25a971d1f4110b32bd3068db3722c80202",
-                "sha256:4373612d59c404baeb7cbd788a18b2b2a8331abcc84c3ba40051fcd18b17a4d5",
-                "sha256:487d63e1454627c8e47dd230025780e91869cfba4c753a74fda196a1f6ad6548",
-                "sha256:48916e459c54c4a70e52745639f1db524542140433599e13911b2f329834276a",
-                "sha256:4922cd707b25e623b902c86188aca466d3620892db76c0bdd7b99a3d5e61d35f",
-                "sha256:55af55e32ae468e9946f741a5d51f9896da6b9bf0bbdd326843fec05c730eb20",
-                "sha256:57e555a9feb4a8460415f1aac331a2dc833b1115284f7ded7278b54afc5bd218",
-                "sha256:5d4b68e216fc65e9fe4f524c177b54964af043dde734807586cf5435af84045c",
-                "sha256:64fda793737bc4037521d4899be780534b9aea552eb673b9833b01f945904c2e",
-                "sha256:6d6169cb3c6c2ad50db5b868db6491a790300ade1ed5d1da29289d73bbe40b56",
-                "sha256:7bcac9a2b4fdbed2c16fa5681356d7121ecabf041f18d97ed5b8e0dd38a80224",
-                "sha256:80b06212075346b5546b0417b9f2bf467fea3bfe7352f781ffc05a8ab24ba14a",
-                "sha256:818014c754cd3dba7229c0f5884396264d51ffb87ec86e927ef0be140bfdb0d2",
-                "sha256:8eb687582ed7cd8c4bdbff3df6c0da443eb89c3c72e6e5dcdd9c81729712791a",
-                "sha256:99f27fefe34c37ba9875f224a8f36e31d744d8083e00f520f133cab79ad5e819",
-                "sha256:9f3e33c28cd39d1b655ed1ba7247133b6f7fc16fa16887b120c0c670e35ce346",
-                "sha256:a8661b2ce9694ca01c529bfa204dbb144b275a31685a075ce123f12331be790b",
-                "sha256:a9da7010cec5a12193d1af9872a00888f396aba3dc79186604a09ea3ee7c029e",
-                "sha256:aedb15f0a5a5949ecb129a82b72b19df97bbbca024081ed2ef88bd5c0a610534",
-                "sha256:b315d709717a99f4b27b59b021e6207c64620790ca3e0bde636a6c7f14618abb",
-                "sha256:ba6f2b3f452e150945d58f4badd92310449876c4c954836cfb1803bdd7b422f0",
-                "sha256:c33d18eb6e6bc36f09d793c0dc58b0211fccc6ae5149b808da4a62660678b156",
-                "sha256:c9a875ce9d7fe32887784274dd533c57909b7b1dcadcc128a2ac21331a9765dd",
-                "sha256:c9e005e9bd57bc987764c32a1bee4364c44fdc11a3cc20a40b93b444984f2b87",
-                "sha256:d2ad4d668a5c0645d281dcd17aff2be3212bc109b33814bbb15c4939f44181cc",
-                "sha256:d950695ae4381ecd856bcaf2b1e866720e4ab9a1498cba61c602e56630ca7195",
-                "sha256:e22dcb48709fc51a7b58a927391b23ab37eb3737a98ac4338e2448bef8559b33",
-                "sha256:e8c6a99be100371dbb046880e7a282152aa5d6127ae01783e37662ef73850d8f",
-                "sha256:e9dc245e3ac69c92ee4c167fbdd7428ec1956d4e754223124991ef29eb57a09d",
-                "sha256:eb687a11f0a7a1839719edd80f41e459cc5366857ecbed383ff376c4e3cc6afd",
-                "sha256:eb9e2a346c5238a30a746893f23a9535e700f8192a68c07c0258e7ece6ff3728",
-                "sha256:ed38b924ce794e505647f7c331b22a693bee1538fdf46b0222c4717b42f744e7",
-                "sha256:f0010c6f9d1a4011e429109fda55a225921e3206e7f62a0c22a35344bfd13cca",
-                "sha256:f0c5d1acbfca6ebdd6b1e3eded8d261affb6ddcf2186205518f1428b8569bb99",
-                "sha256:f10afb1004f102c7868ebfe91c28f4a712227fe4cb24974350ace1f90e1febbf",
-                "sha256:f174135f5609428cc6e1b9090f9268f5c8935fddb1b25ccb8255a2d50de6789e",
-                "sha256:f3ebe6e73c319340830a9b2825d32eb6d8475c1dac020b4f0aa774ee3b898d1c",
-                "sha256:f627688813d0a4140153ff532537fbe4afea5a3dffce1f9deb7f91f848a832b5",
-                "sha256:fd4305f86f53dfd8cd3522269ed7fc34856a8ee3709a5e28b2836b2db9d4cd69"
+                "sha256:00c878c90cb53ccfaae6b8bc18ad05d2036553e6d9d1d9dbcf323bbe83854ca3",
+                "sha256:0104fb5ae2391d46a4cb082abdd5c69ea4eab79d8d44eaaf79f1b1fd806ee4c2",
+                "sha256:06c48159c1abed75c2e721b1715c379fa3200c7784271b3c46df01383b593636",
+                "sha256:0808014eb713677ec1292301ea4c81ad277b6cdf2fdd90fd540af98c0b101d20",
+                "sha256:10dffb601ccfb65262a27233ac273d552ddc4d8ae1bf93b21c94b8511bffe728",
+                "sha256:14cd121ea63ecdae71efa69c15c5543a4b5fbcd0bbe2aad864baca0063cecf27",
+                "sha256:17771976e82e9f94976180f76468546834d22a7cc404b17c22df2a2c81db0c66",
+                "sha256:181dee03b1170ff1969489acf1c26533710231c58f95534e3edac87fff06c443",
+                "sha256:23cfe892bd5dd8941608f93348c0737e369e51c100d03718f108bf1add7bd6d0",
+                "sha256:263cc3d821c4ab2213cbe8cd8b355a7f72a8324577dc865ef98487c1aeee2bc7",
+                "sha256:2756c88cbb94231c7a147402476be2c4df2f6078099a6f4a480d239a8817ae39",
+                "sha256:27c219baf94952ae9d50ec19651a687b826792055353d07648a5695413e0c605",
+                "sha256:2a23af14f408d53d5e6cd4e3d9a24ff9e05906ad574822a10563efcef137979a",
+                "sha256:31fb708d9d7c3f49a60f04cf5b119aeefe5644daba1cd2a0fe389b674fd1de37",
+                "sha256:3415c89f9204ee60cd09b235810be700e993e343a408693e80ce7f6a40108029",
+                "sha256:3773c4d81e6e818df2efbc7dd77325ca0dcb688116050fb2b3011218eda36139",
+                "sha256:3b96a311ac60a3f6be21d2572e46ce67f09abcf4d09344c49274eb9e0bf345fc",
+                "sha256:3f7d084648d77af029acb79a0ff49a0ad7e9d09057a9bf46596dac9514dc07df",
+                "sha256:41d45de54cd277a7878919867c0f08b0cf817605e4eb94093e7516505d3c8d14",
+                "sha256:4238e6dab5d6a8ba812de994bbb0a79bddbdf80994e4ce802b6f6f3142fcc880",
+                "sha256:45db3a33139e9c8f7c09234b5784a5e33d31fd6907800b316decad50af323ff2",
+                "sha256:45e8636704eacc432a206ac7345a5d3d2c62d95a507ec70d62f23cd91770482a",
+                "sha256:4958391dbd6249d7ad855b9ca88fae690783a6be9e86df65865058ed81fc860e",
+                "sha256:4a306fa632e8f0928956a41fa8e1d6243c71e7eb59ffbd165fc0b41e316b2474",
+                "sha256:57e9ac9ccc3101fac9d6014fba037473e4358ef4e89f8e181f8951a2c0162024",
+                "sha256:59888172256cac5629e60e72e86598027aca6bf01fa2465bdb676d37636573e8",
+                "sha256:5e069f72d497312b24fcc02073d70cb989045d1c91cbd53979366077959933e0",
+                "sha256:64d4ec9f448dfe041705426000cc13e34e6e5bb13736e9fd62e34a0b0c41566e",
+                "sha256:6dc2737a3674b3e344847c8686cf29e500584ccad76204efea14f451d4cc669a",
+                "sha256:74fdfdbfdc48d3f47148976f49fab3251e550a8720bebc99bf1483f5bfb5db3e",
+                "sha256:75e4024375654472cc27e91cbe9eaa08567f7fbdf822638be2814ce059f58032",
+                "sha256:786902fb9ba7433aae840e0ed609f45c7bcd4e225ebb9c753aa39725bb3e6ad6",
+                "sha256:8b6c2ea03845c9f501ed1313e78de148cd3f6cad741a75d43a29b43da27f2e1e",
+                "sha256:91d77d2a782be4274da750752bb1650a97bfd8f291022b379bb8e01c66b4e96b",
+                "sha256:91ec59c33514b7c7559a6acda53bbfe1b283949c34fe7440bcf917f96ac0723e",
+                "sha256:920f0d66a896c2d99f0adbb391f990a84091179542c205fa53ce5787aff87954",
+                "sha256:a5263e363c27b653a90078143adb3d076c1a748ec9ecc78ea2fb916f9b861962",
+                "sha256:abb9a20a72ac4e0fdb50dae135ba5e77880518e742077ced47eb1499e29a443c",
+                "sha256:c2051981a968d7de9dd2d7b87bcb9c939c74a34626a6e2f8181455dd49ed69e4",
+                "sha256:c21c9e3896c23007803a875460fb786118f0cdd4434359577ea25eb556e34c55",
+                "sha256:c2502a1a03b6312837279c8c1bd3ebedf6c12c4228ddbad40912d671ccc8a962",
+                "sha256:d4d692a89c5cf08a8557fdeb329b82e7bf609aadfaed6c0d79f5a449a3c7c023",
+                "sha256:da5db4e883f1ce37f55c667e5c0de439df76ac4cb55964655906306918e7363c",
+                "sha256:e7022a66d9b55e93e1a845d8c9eba2a1bebd4966cd8bfc25d9cd07d515b33fa6",
+                "sha256:ef1f279350da2c586a69d32fc8733092fd32cc8ac95139a00377841f59a3f8d8",
+                "sha256:f54a64f8b0c8ff0b64d18aa76675262e1700f3995182267998c31ae974fbc382",
+                "sha256:f5c7150ad32ba43a07c4479f40241756145a1f03b43480e058cfd862bf5041c7",
+                "sha256:f6f824dc3bce0edab5f427efcfb1d63ee75b6fcb7282900ccaf925be84efb0fc",
+                "sha256:fd8a250edc26254fe5b33be00402e6d287f562b6a5b2152dec302fa15bb3e997",
+                "sha256:ffaa5c925128e29efbde7301d8ecaf35c8c60ffbcd6a1ffd3a552177c8e5e796"
             ],
-            "version": "==1.14.6"
+            "version": "==1.15.0"
         },
         "charset-normalizer": {
             "hashes": [
-                "sha256:5d209c0a931f215cee683b6445e2d77677e7e75e159f78def0db09d68fafcaa6",
-                "sha256:5ec46d183433dcbd0ab716f2d7f29d8dee50505b3fdb40c6b985c7c4f5a3591f"
+                "sha256:e019de665e2bcf9c2b64e2e5aa025fa991da8720daa3c1138cadd2fd1856aed0",
+                "sha256:f7af805c321bfa1ce6714c51f254e0d5bb5e5834039bc17db7ebe3a4cec9492b"
             ],
             "markers": "python_version >= '3'",
-            "version": "==2.0.6"
+            "version": "==2.0.7"
         },
         "cryptography": {
             "hashes": [
-                "sha256:0a7dcbcd3f1913f664aca35d47c1331fce738d44ec34b7be8b9d332151b0b01e",
-                "sha256:1eb7bb0df6f6f583dd8e054689def236255161ebbcf62b226454ab9ec663746b",
-                "sha256:21ca464b3a4b8d8e86ba0ee5045e103a1fcfac3b39319727bc0fc58c09c6aff7",
-                "sha256:34dae04a0dce5730d8eb7894eab617d8a70d0c97da76b905de9efb7128ad7085",
-                "sha256:3520667fda779eb788ea00080124875be18f2d8f0848ec00733c0ec3bb8219fc",
-                "sha256:3c4129fc3fdc0fa8e40861b5ac0c673315b3c902bbdc05fc176764815b43dd1d",
-                "sha256:3fa3a7ccf96e826affdf1a0a9432be74dc73423125c8f96a909e3835a5ef194a",
-                "sha256:5b0fbfae7ff7febdb74b574055c7466da334a5371f253732d7e2e7525d570498",
-                "sha256:695104a9223a7239d155d7627ad912953b540929ef97ae0c34c7b8bf30857e89",
-                "sha256:8695456444f277af73a4877db9fc979849cd3ee74c198d04fc0776ebc3db52b9",
-                "sha256:94cc5ed4ceaefcbe5bf38c8fba6a21fc1d365bb8fb826ea1688e3370b2e24a1c",
-                "sha256:94fff993ee9bc1b2440d3b7243d488c6a3d9724cc2b09cdb297f6a886d040ef7",
-                "sha256:9965c46c674ba8cc572bc09a03f4c649292ee73e1b683adb1ce81e82e9a6a0fb",
-                "sha256:a00cf305f07b26c351d8d4e1af84ad7501eca8a342dedf24a7acb0e7b7406e14",
-                "sha256:a305600e7a6b7b855cd798e00278161b681ad6e9b7eca94c721d5f588ab212af",
-                "sha256:cd65b60cfe004790c795cc35f272e41a3df4631e2fb6b35aa7ac6ef2859d554e",
-                "sha256:d2a6e5ef66503da51d2110edf6c403dc6b494cc0082f85db12f54e9c5d4c3ec5",
-                "sha256:d9ec0e67a14f9d1d48dd87a2531009a9b251c02ea42851c060b25c782516ff06",
-                "sha256:f44d141b8c4ea5eb4dbc9b3ad992d45580c1d22bf5e24363f2fbf50c2d7ae8a7"
+                "sha256:07bb7fbfb5de0980590ddfc7f13081520def06dc9ed214000ad4372fb4e3c7f6",
+                "sha256:18d90f4711bf63e2fb21e8c8e51ed8189438e6b35a6d996201ebd98a26abbbe6",
+                "sha256:1ed82abf16df40a60942a8c211251ae72858b25b7421ce2497c2eb7a1cee817c",
+                "sha256:22a38e96118a4ce3b97509443feace1d1011d0571fae81fc3ad35f25ba3ea999",
+                "sha256:2d69645f535f4b2c722cfb07a8eab916265545b3475fdb34e0be2f4ee8b0b15e",
+                "sha256:4a2d0e0acc20ede0f06ef7aa58546eee96d2592c00f450c9acb89c5879b61992",
+                "sha256:54b2605e5475944e2213258e0ab8696f4f357a31371e538ef21e8d61c843c28d",
+                "sha256:7075b304cd567694dc692ffc9747f3e9cb393cc4aa4fb7b9f3abd6f5c4e43588",
+                "sha256:7b7ceeff114c31f285528ba8b390d3e9cfa2da17b56f11d366769a807f17cbaa",
+                "sha256:7eba2cebca600a7806b893cb1d541a6e910afa87e97acf2021a22b32da1df52d",
+                "sha256:928185a6d1ccdb816e883f56ebe92e975a262d31cc536429041921f8cb5a62fd",
+                "sha256:9933f28f70d0517686bd7de36166dda42094eac49415459d9bdf5e7df3e0086d",
+                "sha256:a688ebcd08250eab5bb5bca318cc05a8c66de5e4171a65ca51db6bd753ff8953",
+                "sha256:abb5a361d2585bb95012a19ed9b2c8f412c5d723a9836418fab7aaa0243e67d2",
+                "sha256:c10c797ac89c746e488d2ee92bd4abd593615694ee17b2500578b63cad6b93a8",
+                "sha256:ced40344e811d6abba00295ced98c01aecf0c2de39481792d87af4fa58b7b4d6",
+                "sha256:d57e0cdc1b44b6cdf8af1d01807db06886f10177469312fbde8f44ccbb284bc9",
+                "sha256:d99915d6ab265c22873f1b4d6ea5ef462ef797b4140be4c9d8b179915e0985c6",
+                "sha256:eb80e8a1f91e4b7ef8b33041591e6d89b2b8e122d787e87eeb2b08da71bb16ad",
+                "sha256:ebeddd119f526bcf323a89f853afb12e225902a24d29b55fe18dd6fcb2838a76"
             ],
-            "version": "==3.4.8"
+            "version": "==35.0.0"
         },
         "execnet": {
             "hashes": [
@@ -142,11 +156,11 @@
         },
         "idna": {
             "hashes": [
-                "sha256:14475042e284991034cb48e06f6851428fb14c4dc953acd9be9a5e95c7b6dd7a",
-                "sha256:467fbad99067910785144ce333826c71fb0e63a425657295239737f7ecd125f3"
+                "sha256:84d9dd047ffa80596e0f246e2eab0b391788b0503584e8945f2368256d2735ff",
+                "sha256:9d643ff0a55b762d5cdb124b8eaa99c66322e2157b69160bc32796e824360e6d"
             ],
             "markers": "python_version >= '3'",
-            "version": "==3.2"
+            "version": "==3.3"
         },
         "iniconfig": {
             "hashes": [
@@ -207,11 +221,11 @@
                 "crypto"
             ],
             "hashes": [
-                "sha256:934d73fbba91b0483d3857d1aff50e96b2a892384ee2c17417ed3203f173fca1",
-                "sha256:fba44e7898bbca160a2b2b501f492824fc8382485d3a6f11ba5d0c1937ce6130"
+                "sha256:b888b4d56f06f6dcd777210c334e69c737be74755d3e5e9ee3fe67dc18a0ee41",
+                "sha256:e0c4bb8d9f0af0c7f5b1ec4c5036309617d03d56932877f2f7a0beeb5318322f"
             ],
             "index": "pypi",
-            "version": "==2.1.0"
+            "version": "==2.3.0"
         },
         "pyparsing": {
             "hashes": [
@@ -272,21 +286,21 @@
         },
         "urllib3": {
             "hashes": [
-                "sha256:39fb8672126159acb139a7718dd10806104dec1e2f0f6c88aab05d17df10c8d4",
-                "sha256:f57b4c16c62fa2760b7e3d97c35b255512fb6b59a259730f36ba32ce9f8e342f"
+                "sha256:4987c65554f7a2dbf30c18fd48778ef124af6fab771a377103da0585e2336ece",
+                "sha256:c4fdf4019605b6e5423637e01bc9fe4daef873709a7973e195ceba0a62bbc844"
             ],
             "markers": "python_version >= '2.7' and python_version not in '3.0, 3.1, 3.2, 3.3, 3.4' and python_version < '4'",
-            "version": "==1.26.6"
+            "version": "==1.26.7"
         }
     },
     "develop": {
         "flake8": {
             "hashes": [
-                "sha256:07528381786f2a6237b061f6e96610a4167b226cb926e2aa2b6b1d78057c576b",
-                "sha256:bf8fd333346d844f616e8d47905ef3a3384edae6b4e9beb0c5101e25e3110907"
+                "sha256:479b1304f72536a55948cb40a32dce8bb0ffe3501e26eaf292c7e60eb5e0428d",
+                "sha256:806e034dda44114815e23c16ef92f95c91e4c71100ff52813adf7132a6ad870d"
             ],
             "index": "pypi",
-            "version": "==3.9.2"
+            "version": "==4.0.1"
         },
         "mccabe": {
             "hashes": [
@@ -333,19 +347,19 @@
         },
         "pycodestyle": {
             "hashes": [
-                "sha256:514f76d918fcc0b55c6680472f0a37970994e07bbb80725808c17089be302068",
-                "sha256:c389c1d06bf7904078ca03399a4816f974a1d590090fecea0c63ec26ebaf1cef"
+                "sha256:720f8b39dde8b293825e7ff02c475f3077124006db4f440dcbc9a20b76548a20",
+                "sha256:eddd5847ef438ea1c7870ca7eb78a9d47ce0cdb4851a5523949f2601d0cbbe7f"
             ],
-            "markers": "python_version >= '2.7' and python_version not in '3.0, 3.1, 3.2, 3.3'",
-            "version": "==2.7.0"
+            "markers": "python_version >= '2.7' and python_version not in '3.0, 3.1, 3.2, 3.3, 3.4'",
+            "version": "==2.8.0"
         },
         "pyflakes": {
             "hashes": [
-                "sha256:7893783d01b8a89811dd72d7dfd4d84ff098e5eed95cfa8905b22bbffe52efc3",
-                "sha256:f5bc8ecabc05bb9d291eb5203d6810b49040f6ff446a756326104746cc00c1db"
+                "sha256:05a85c2872edf37a4ed30b0cce2f6093e1d0581f8c19d7393122da7e25b2b24c",
+                "sha256:3bb3a3f256f4b7968c9c788781e4ff07dce46bdf12339dcda61053375426ee2e"
             ],
             "markers": "python_version >= '2.7' and python_version not in '3.0, 3.1, 3.2, 3.3'",
-            "version": "==2.3.1"
+            "version": "==2.4.0"
         },
         "toml": {
             "hashes": [

test_runner/README.md

@@ -53,8 +53,8 @@ Useful environment variables:
 should go.
 `TEST_SHARED_FIXTURES`: Try to re-use a single pageserver for all the tests.
 
-Let stdout and stderr go to the terminal instead of capturing them:
-`pytest -s ...`
+Let stdout, stderr and `INFO` log messages go to the terminal instead of capturing them:
+`pytest -s --log-cli-level=INFO ...`
 (Note many tests capture subprocess outputs separately, so this may not
 show much.)
 
@@ -95,11 +95,13 @@ Python destructors, e.g. `__del__()` aren't recommended for cleanup.
 
 ### Code quality
 
+We force code formatting via yapf:
+
+1. Install `yapf` and other tools (`flake8`, `mypy`) with `pipenv install --dev`.
+1. Reformat all your code by running `pipenv run yapf -ri .` in the `test_runner/` directory.
+
 Before submitting a patch, please consider:
 
 * Writing a couple of docstrings to clarify the reasoning behind a new test.
 * Running `flake8` (or a linter of your choice, e.g. `pycodestyle`) and fixing possible defects, if any.
-* Formatting the code with `yapf -r -i .` (TODO: implement an opt-in pre-commit hook for that).
 * (Optional) Typechecking the code with `mypy .`. Currently this mostly affects `fixtures/zenith_fixtures.py`.
-
-The tools can be installed with `pipenv install --dev`.

test_runner/batch_others/test_auth.py

@@ -1,4 +1,3 @@
-
 from contextlib import closing
 from typing import Iterator
 from uuid import uuid4
@@ -6,7 +5,6 @@ import psycopg2
 from fixtures.zenith_fixtures import PortDistributor, Postgres, ZenithCli, ZenithPageserver, PgBin
 import pytest
 
-
 pytest_plugins = ("fixtures.zenith_fixtures")
 
 
@@ -35,7 +33,9 @@ def test_pageserver_auth(pageserver_auth_enabled: ZenithPageserver):
     ps.safe_psql(f"tenant_create {uuid4().hex}", password=management_token)
 
     # fail to create tenant using tenant token
-    with pytest.raises(psycopg2.DatabaseError, match='Attempt to access management api with tenant scope. Permission denied'):
+    with pytest.raises(
+            psycopg2.DatabaseError,
+            match='Attempt to access management api with tenant scope. Permission denied'):
         ps.safe_psql(f"tenant_create {uuid4().hex}", password=tenant_token)
 
 
@@ -60,14 +60,14 @@ def test_compute_auth_to_pageserver(
         wa_factory.start_n_new(3, management_token)
 
     with Postgres(
-        zenith_cli=zenith_cli,
-        repo_dir=repo_dir,
-        pg_bin=pg_bin,
-        tenant_id=ps.initial_tenant,
-        port=port_distributor.get_port(),
+            zenith_cli=zenith_cli,
+            repo_dir=repo_dir,
+            pg_bin=pg_bin,
+            tenant_id=ps.initial_tenant,
+            port=port_distributor.get_port(),
     ).create_start(
-        branch,
-        wal_acceptors=wa_factory.get_connstrs() if with_wal_acceptors else None,
+            branch,
+            wal_acceptors=wa_factory.get_connstrs() if with_wal_acceptors else None,
     ) as pg:
         with closing(pg.connect()) as conn:
             with conn.cursor() as cur:

test_runner/batch_others/test_branch_behind.py

@@ -86,7 +86,10 @@ def test_branch_behind(zenith_cli, pageserver: ZenithPageserver, postgres: Postg
     assert cur.fetchone() == (1, )
 
     # branch at pre-initdb lsn
+    #
+    # FIXME: This works currently, but probably shouldn't be allowed
     try:
         zenith_cli.run(["branch", "test_branch_preinitdb", "test_branch_behind@0/42"])
+        # FIXME: assert false, "branch with invalid LSN should have failed"
     except subprocess.CalledProcessError:
         log.info("Branch creation with pre-initdb LSN failed (as expected)")

test_runner/batch_others/test_clog_truncate.py

@@ -18,9 +18,12 @@ def test_clog_truncate(zenith_cli, pageserver: ZenithPageserver, postgres: Postg
 
     # set agressive autovacuum to make sure that truncation will happen
     config = [
-        'autovacuum_max_workers=10', 'autovacuum_vacuum_threshold=0',
-        'autovacuum_vacuum_insert_threshold=0', 'autovacuum_vacuum_cost_delay=0',
-        'autovacuum_vacuum_cost_limit=10000', 'autovacuum_naptime =1s',
+        'autovacuum_max_workers=10',
+        'autovacuum_vacuum_threshold=0',
+        'autovacuum_vacuum_insert_threshold=0',
+        'autovacuum_vacuum_cost_delay=0',
+        'autovacuum_vacuum_cost_limit=10000',
+        'autovacuum_naptime =1s',
         'autovacuum_freeze_max_age=100000'
     ]
 

test_runner/batch_others/test_createdropdb.py

@@ -41,6 +41,7 @@ def test_createdb(
     for db in (pg, pg2):
         db.connect(dbname='foodb').close()
 
+
 #
 # Test DROP DATABASE
 #
@@ -49,7 +50,7 @@ def test_dropdb(
     pageserver: ZenithPageserver,
     postgres: PostgresFactory,
     pg_bin,
-    test_output_dir
+    test_output_dir,
 ):
     zenith_cli.run(["branch", "test_dropdb", "empty"])
 
@@ -66,7 +67,6 @@ def test_dropdb(
             cur.execute("SELECT oid FROM pg_database WHERE datname='foodb';")
             dboid = cur.fetchone()[0]
 
-
     with closing(pg.connect()) as conn:
         with conn.cursor() as cur:
             cur.execute('DROP DATABASE foodb')
@@ -76,7 +76,6 @@ def test_dropdb(
             cur.execute('SELECT pg_current_wal_insert_lsn()')
             lsn_after_drop = cur.fetchone()[0]
 
-
     # Create two branches before and after database drop.
     zenith_cli.run(["branch", "test_before_dropdb", "test_dropdb@" + lsn_before_drop])
     pg_before = postgres.create_start('test_before_dropdb')

test_runner/batch_others/test_multixact.py

@@ -10,8 +10,12 @@ pytest_plugins = ("fixtures.zenith_fixtures")
 # it only checks next_multixact_id field in restored pg_control,
 # since we don't have functions to check multixact internals.
 #
-def test_multixact(pageserver: ZenithPageserver, postgres: PostgresFactory,
-                    pg_bin, zenith_cli, base_dir, test_output_dir):
+def test_multixact(pageserver: ZenithPageserver,
+                   postgres: PostgresFactory,
+                   pg_bin,
+                   zenith_cli,
+                   base_dir,
+                   test_output_dir):
     # Create a branch for us
     zenith_cli.run(["branch", "test_multixact", "empty"])
     pg = postgres.create_start('test_multixact')

test_runner/batch_others/test_old_request_lsn.py

@@ -5,6 +5,7 @@ from fixtures.log_helper import log
 
 pytest_plugins = ("fixtures.zenith_fixtures")
 
+
 #
 # Test where Postgres generates a lot of WAL, and it's garbage collected away, but
 # no pages are evicted so that Postgres uses an old LSN in a GetPage request.
@@ -15,7 +16,10 @@ pytest_plugins = ("fixtures.zenith_fixtures")
 # just a hint that the page hasn't been modified since that LSN, and the page
 # server should return the latest page version regardless of the LSN.
 #
-def test_old_request_lsn(zenith_cli, pageserver: ZenithPageserver, postgres: PostgresFactory, pg_bin):
+def test_old_request_lsn(zenith_cli,
+                         pageserver: ZenithPageserver,
+                         postgres: PostgresFactory,
+                         pg_bin):
     # Create a branch for us
     zenith_cli.run(["branch", "test_old_request_lsn", "empty"])
     pg = postgres.create_start('test_old_request_lsn')
@@ -47,20 +51,20 @@ 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]}');
+    log.info(f'shared_buffers is {row[0]}, table size {row[1]}')
     assert int(row[0]) < int(row[1])
 
-    cur.execute('VACUUM foo');
+    cur.execute('VACUUM foo')
 
     # Make a lot of updates on a single row, generating a lot of WAL. Trigger
     # garbage collections so that the page server will remove old page versions.
     for i in range(10):
         pscur.execute(f"do_gc {pageserver.initial_tenant} {timeline} 0")
         for j in range(100):
-            cur.execute('UPDATE foo SET val = val + 1 WHERE id = 1;');
+            cur.execute('UPDATE foo SET val = val + 1 WHERE id = 1;')
 
     # All (or at least most of) the updates should've been on the same page, so
     # that we haven't had to evict any dirty pages for a long time. Now run
     # a query that sends GetPage@LSN requests with the old LSN.
-    cur.execute("SELECT COUNT(*), SUM(val) FROM foo");
+    cur.execute("SELECT COUNT(*), SUM(val) FROM foo")
     assert cur.fetchone() == (100000, 101000)

test_runner/batch_others/test_pageserver_api.py

@@ -63,7 +63,8 @@ def test_tenant_list_psql(pageserver: ZenithPageserver, zenith_cli):
     cur = conn.cursor()
 
     # check same tenant cannot be created twice
-    with pytest.raises(psycopg2.DatabaseError, match=f'tenant {pageserver.initial_tenant} already exists'):
+    with pytest.raises(psycopg2.DatabaseError,
+                       match=f'tenant {pageserver.initial_tenant} already exists'):
         cur.execute(f'tenant_create {pageserver.initial_tenant}')
 
     # create one more tenant
@@ -102,5 +103,6 @@ def test_pageserver_http_api_client(pageserver: ZenithPageserver):
 
 
 def test_pageserver_http_api_client_auth_enabled(pageserver_auth_enabled: ZenithPageserver):
-    client = pageserver_auth_enabled.http_client(auth_token=pageserver_auth_enabled.auth_keys.generate_management_token())
+    client = pageserver_auth_enabled.http_client(
+        auth_token=pageserver_auth_enabled.auth_keys.generate_management_token())
     check_client(client, pageserver_auth_enabled.initial_tenant)

test_runner/batch_others/test_pageserver_restart.py

@@ -9,17 +9,20 @@ from fixtures.log_helper import log
 
 pytest_plugins = ("fixtures.zenith_fixtures")
 
+
 # Check that dead minority doesn't prevent the commits: execute insert n_inserts
 # times, with fault_probability chance of getting a wal acceptor down or up
 # along the way. 2 of 3 are always alive, so the work keeps going.
-def test_pageserver_restart(zenith_cli, pageserver: ZenithPageserver, postgres: PostgresFactory, wa_factory: WalAcceptorFactory):
+def test_pageserver_restart(zenith_cli,
+                            pageserver: ZenithPageserver,
+                            postgres: PostgresFactory,
+                            wa_factory: WalAcceptorFactory):
 
     # One safekeeper is enough for this test.
     wa_factory.start_n_new(1)
 
     zenith_cli.run(["branch", "test_pageserver_restart", "empty"])
-    pg = postgres.create_start('test_pageserver_restart',
-                               wal_acceptors=wa_factory.get_connstrs())
+    pg = postgres.create_start('test_pageserver_restart', wal_acceptors=wa_factory.get_connstrs())
 
     pg_conn = pg.connect()
     cur = pg_conn.cursor()
@@ -41,14 +44,14 @@ 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]}");
+    log.info(f"shared_buffers is {row[0]}, table size {row[1]}")
     assert int(row[0]) < int(row[1])
 
     # Stop and restart pageserver. This is a more or less graceful shutdown, although
     # the page server doesn't currently have a shutdown routine so there's no difference
     # between stopping and crashing.
-    pageserver.stop();
-    pageserver.start();
+    pageserver.stop()
+    pageserver.start()
 
     # Stopping the pageserver breaks the connection from the postgres backend to
     # the page server, and causes the next query on the connection to fail. Start a new
@@ -62,6 +65,5 @@ def test_pageserver_restart(zenith_cli, pageserver: ZenithPageserver, postgres:
     assert cur.fetchone() == (100000, )
 
     # Stop the page server by force, and restart it
-    pageserver.stop();
-    pageserver.start();
-
+    pageserver.stop()
+    pageserver.start()

test_runner/batch_others/test_readonly_node.py

@@ -0,0 +1,89 @@
+import subprocess
+from fixtures.zenith_fixtures import PostgresFactory, ZenithPageserver
+
+pytest_plugins = ("fixtures.zenith_fixtures")
+
+
+#
+# Create read-only compute nodes, anchored at historical points in time.
+#
+# This is very similar to the 'test_branch_behind' test, but instead of
+# creating branches, creates read-only nodes.
+#
+def test_readonly_node(zenith_cli, pageserver: ZenithPageserver, postgres: PostgresFactory, pg_bin):
+    zenith_cli.run(["branch", "test_readonly_node", "empty"])
+
+    pgmain = postgres.create_start('test_readonly_node')
+    print("postgres is running on 'test_readonly_node' branch")
+
+    main_pg_conn = pgmain.connect()
+    main_cur = main_pg_conn.cursor()
+
+    # Create table, and insert the first 100 rows
+    main_cur.execute('CREATE TABLE foo (t text)')
+    main_cur.execute('''
+        INSERT INTO foo
+            SELECT 'long string to consume some space' || g
+            FROM generate_series(1, 100) g
+    ''')
+    main_cur.execute('SELECT pg_current_wal_insert_lsn()')
+    lsn_a = main_cur.fetchone()[0]
+    print('LSN after 100 rows: ' + lsn_a)
+
+    # Insert some more rows. (This generates enough WAL to fill a few segments.)
+    main_cur.execute('''
+        INSERT INTO foo
+            SELECT 'long string to consume some space' || g
+            FROM generate_series(1, 200000) g
+    ''')
+    main_cur.execute('SELECT pg_current_wal_insert_lsn()')
+    lsn_b = main_cur.fetchone()[0]
+    print('LSN after 200100 rows: ' + lsn_b)
+
+    # Insert many more rows. This generates enough WAL to fill a few segments.
+    main_cur.execute('''
+        INSERT INTO foo
+            SELECT 'long string to consume some space' || g
+            FROM generate_series(1, 200000) g
+    ''')
+
+    main_cur.execute('SELECT pg_current_wal_insert_lsn()')
+    lsn_c = main_cur.fetchone()[0]
+    print('LSN after 400100 rows: ' + lsn_c)
+
+    # Create first read-only node at the point where only 100 rows were inserted
+    pg_hundred = postgres.create_start("test_readonly_node_hundred",
+                                       branch=f'test_readonly_node@{lsn_a}')
+
+    # And another at the point where 200100 rows were inserted
+    pg_more = postgres.create_start("test_readonly_node_more", branch=f'test_readonly_node@{lsn_b}')
+
+    # On the 'hundred' node, we should see only 100 rows
+    hundred_pg_conn = pg_hundred.connect()
+    hundred_cur = hundred_pg_conn.cursor()
+    hundred_cur.execute('SELECT count(*) FROM foo')
+    assert hundred_cur.fetchone() == (100, )
+
+    # On the 'more' node, we should see 100200 rows
+    more_pg_conn = pg_more.connect()
+    more_cur = more_pg_conn.cursor()
+    more_cur.execute('SELECT count(*) FROM foo')
+    assert more_cur.fetchone() == (200100, )
+
+    # All the rows are visible on the main branch
+    main_cur.execute('SELECT count(*) FROM foo')
+    assert main_cur.fetchone() == (400100, )
+
+    # Check creating a node at segment boundary
+    pg = postgres.create_start("test_branch_segment_boundary",
+                               branch="test_readonly_node@0/3000000")
+    cur = pg.connect().cursor()
+    cur.execute('SELECT 1')
+    assert cur.fetchone() == (1, )
+
+    # Create node at pre-initdb lsn
+    try:
+        zenith_cli.run(["pg", "start", "test_branch_preinitdb", "test_readonly_node@0/42"])
+        assert false, "compute node startup with invalid LSN should have failed"
+    except Exception:
+        print("Node creation with pre-initdb LSN failed (as expected)")

test_runner/batch_others/test_restart_compute.py

@@ -12,13 +12,13 @@ pytest_plugins = ("fixtures.zenith_fixtures")
 #
 @pytest.mark.parametrize('with_wal_acceptors', [False, True])
 def test_restart_compute(
-        zenith_cli,
-        pageserver: ZenithPageserver,
-        postgres: PostgresFactory,
-        pg_bin,
-        wa_factory,
-        with_wal_acceptors: bool,
-    ):
+    zenith_cli,
+    pageserver: ZenithPageserver,
+    postgres: PostgresFactory,
+    pg_bin,
+    wa_factory,
+    with_wal_acceptors: bool,
+):
     wal_acceptor_connstrs = None
     zenith_cli.run(["branch", "test_restart_compute", "empty"])
 
@@ -26,8 +26,7 @@ def test_restart_compute(
         wa_factory.start_n_new(3)
         wal_acceptor_connstrs = wa_factory.get_connstrs()
 
-    pg = postgres.create_start('test_restart_compute',
-                               wal_acceptors=wal_acceptor_connstrs)
+    pg = postgres.create_start('test_restart_compute', wal_acceptors=wal_acceptor_connstrs)
     log.info("postgres is running on 'test_restart_compute' branch")
 
     with closing(pg.connect()) as conn:
@@ -40,9 +39,7 @@ def test_restart_compute(
             log.info(f"res = {r}")
 
     # Remove data directory and restart
-    pg.stop_and_destroy().create_start('test_restart_compute',
-                                       wal_acceptors=wal_acceptor_connstrs)
-
+    pg.stop_and_destroy().create_start('test_restart_compute', wal_acceptors=wal_acceptor_connstrs)
 
     with closing(pg.connect()) as conn:
         with conn.cursor() as cur:
@@ -61,8 +58,7 @@ def test_restart_compute(
             log.info(f"res = {r}")
 
     # Again remove data directory and restart
-    pg.stop_and_destroy().create_start('test_restart_compute',
-                                       wal_acceptors=wal_acceptor_connstrs)
+    pg.stop_and_destroy().create_start('test_restart_compute', wal_acceptors=wal_acceptor_connstrs)
 
     # That select causes lots of FPI's and increases probability of wakeepers
     # lagging behind after query completion
@@ -76,8 +72,7 @@ def test_restart_compute(
             log.info(f"res = {r}")
 
     # And again remove data directory and restart
-    pg.stop_and_destroy().create_start('test_restart_compute',
-                                       wal_acceptors=wal_acceptor_connstrs)
+    pg.stop_and_destroy().create_start('test_restart_compute', wal_acceptors=wal_acceptor_connstrs)
 
     with closing(pg.connect()) as conn:
         with conn.cursor() as cur:

test_runner/batch_others/test_snapfiles_gc.py

@@ -5,10 +5,15 @@ from fixtures.log_helper import log
 
 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))
+    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))
 
 
 #
@@ -24,7 +29,7 @@ def test_layerfiles_gc(zenith_cli, pageserver, postgres, pg_bin):
     with closing(pg.connect()) as conn:
         with conn.cursor() as cur:
             with closing(pageserver.connect()) as psconn:
-                with psconn.cursor(cursor_factory = psycopg2.extras.DictCursor) as pscur:
+                with psconn.cursor(cursor_factory=psycopg2.extras.DictCursor) as pscur:
 
                     # Get the timeline ID of our branch. We need it for the 'do_gc' command
                     cur.execute("SHOW zenith.zenith_timeline")
@@ -34,9 +39,9 @@ def test_layerfiles_gc(zenith_cli, pageserver, postgres, pg_bin):
                     cur.execute("CREATE TABLE foo(x integer)")
                     cur.execute("INSERT INTO foo VALUES (1)")
 
-                    cur.execute("select relfilenode from pg_class where oid = 'foo'::regclass");
-                    row = cur.fetchone();
-                    log.info(f"relfilenode is {row[0]}");
+                    cur.execute("select relfilenode from pg_class where oid = 'foo'::regclass")
+                    row = cur.fetchone()
+                    log.info(f"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
@@ -54,9 +59,10 @@ def test_layerfiles_gc(zenith_cli, pageserver, postgres, pg_bin):
                     log.info("Running GC before test")
                     pscur.execute(f"do_gc {pageserver.initial_tenant} {timeline} 0")
                     row = pscur.fetchone()
-                    print_gc_result(row);
+                    print_gc_result(row)
                     # remember the number of files
-                    layer_relfiles_remain = row['layer_relfiles_total'] - row['layer_relfiles_removed']
+                    layer_relfiles_remain = (row['layer_relfiles_total'] -
+                                             row['layer_relfiles_removed'])
                     assert layer_relfiles_remain > 0
 
                     # Insert a row and run GC. Checkpoint should freeze the layer
@@ -66,7 +72,7 @@ def test_layerfiles_gc(zenith_cli, pageserver, postgres, pg_bin):
                     cur.execute("INSERT INTO foo VALUES (1)")
                     pscur.execute(f"do_gc {pageserver.initial_tenant} {timeline} 0")
                     row = pscur.fetchone()
-                    print_gc_result(row);
+                    print_gc_result(row)
                     assert row['layer_relfiles_total'] == layer_relfiles_remain + 2
                     assert row['layer_relfiles_removed'] == 2
                     assert row['layer_relfiles_dropped'] == 0
@@ -80,7 +86,7 @@ def test_layerfiles_gc(zenith_cli, pageserver, postgres, pg_bin):
 
                     pscur.execute(f"do_gc {pageserver.initial_tenant} {timeline} 0")
                     row = pscur.fetchone()
-                    print_gc_result(row);
+                    print_gc_result(row)
                     assert row['layer_relfiles_total'] == layer_relfiles_remain + 2
                     assert row['layer_relfiles_removed'] == 2
                     assert row['layer_relfiles_dropped'] == 0
@@ -92,7 +98,7 @@ def test_layerfiles_gc(zenith_cli, pageserver, postgres, pg_bin):
 
                     pscur.execute(f"do_gc {pageserver.initial_tenant} {timeline} 0")
                     row = pscur.fetchone()
-                    print_gc_result(row);
+                    print_gc_result(row)
                     assert row['layer_relfiles_total'] == layer_relfiles_remain + 2
                     assert row['layer_relfiles_removed'] == 2
                     assert row['layer_relfiles_dropped'] == 0
@@ -101,7 +107,7 @@ def test_layerfiles_gc(zenith_cli, pageserver, postgres, pg_bin):
                     log.info("Run GC again, with nothing to do")
                     pscur.execute(f"do_gc {pageserver.initial_tenant} {timeline} 0")
                     row = pscur.fetchone()
-                    print_gc_result(row);
+                    print_gc_result(row)
                     assert row['layer_relfiles_total'] == layer_relfiles_remain
                     assert row['layer_relfiles_removed'] == 0
                     assert row['layer_relfiles_dropped'] == 0
@@ -109,12 +115,12 @@ 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");
+                    log.info("Drop table and run GC again")
                     cur.execute("DROP TABLE foo")
 
                     pscur.execute(f"do_gc {pageserver.initial_tenant} {timeline} 0")
                     row = pscur.fetchone()
-                    print_gc_result(row);
+                    print_gc_result(row)
 
                     # We still cannot remove the latest layers
                     # because they serve as tombstones for earlier layers.

test_runner/batch_others/test_tenants.py

@@ -21,18 +21,30 @@ def test_tenants_normal_work(
     tenant_1 = tenant_factory.create()
     tenant_2 = tenant_factory.create()
 
-    zenith_cli.run(["branch", f"test_tenants_normal_work_with_wal_acceptors{with_wal_acceptors}", "main", f"--tenantid={tenant_1}"])
-    zenith_cli.run(["branch", f"test_tenants_normal_work_with_wal_acceptors{with_wal_acceptors}", "main", f"--tenantid={tenant_2}"])
+    zenith_cli.run([
+        "branch",
+        f"test_tenants_normal_work_with_wal_acceptors{with_wal_acceptors}",
+        "main",
+        f"--tenantid={tenant_1}"
+    ])
+    zenith_cli.run([
+        "branch",
+        f"test_tenants_normal_work_with_wal_acceptors{with_wal_acceptors}",
+        "main",
+        f"--tenantid={tenant_2}"
+    ])
     if with_wal_acceptors:
         wa_factory.start_n_new(3)
 
     pg_tenant1 = postgres.create_start(
         f"test_tenants_normal_work_with_wal_acceptors{with_wal_acceptors}",
+        None,  # branch name, None means same as node name
         tenant_1,
         wal_acceptors=wa_factory.get_connstrs() if with_wal_acceptors else None,
     )
     pg_tenant2 = postgres.create_start(
         f"test_tenants_normal_work_with_wal_acceptors{with_wal_acceptors}",
+        None,  # branch name, None means same as node name
         tenant_2,
         wal_acceptors=wa_factory.get_connstrs() if with_wal_acceptors else None,
     )
@@ -45,4 +57,4 @@ def test_tenants_normal_work(
                 cur.execute("CREATE TABLE t(key int primary key, value text)")
                 cur.execute("INSERT INTO t SELECT generate_series(1,100000), 'payload'")
                 cur.execute("SELECT sum(key) FROM t")
-                assert cur.fetchone() == (5000050000,)
+                assert cur.fetchone() == (5000050000, )

test_runner/batch_others/test_timeline_size.py

@@ -4,9 +4,8 @@ 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
-):
+
+def test_timeline_size(zenith_cli, pageserver: ZenithPageserver, postgres: PostgresFactory, pg_bin):
     # Branch at the point where only 100 rows were inserted
     zenith_cli.run(["branch", "test_timeline_size", "empty"])
 
@@ -23,13 +22,11 @@ def test_timeline_size(
 
             # Create table, and insert the first 100 rows
             cur.execute("CREATE TABLE foo (t text)")
-            cur.execute(
-                """
+            cur.execute("""
                 INSERT INTO foo
                     SELECT 'long string to consume some space' || g
                     FROM generate_series(1, 10) g
-            """
-            )
+            """)
 
             res = client.branch_detail(UUID(pageserver.initial_tenant), "test_timeline_size")
             assert res["current_logical_size"] == res["current_logical_size_non_incremental"]

test_runner/batch_others/test_twophase.py

@@ -9,7 +9,10 @@ pytest_plugins = ("fixtures.zenith_fixtures")
 #
 # Test branching, when a transaction is in prepared state
 #
-def test_twophase(zenith_cli, pageserver: ZenithPageserver, postgres: PostgresFactory, pg_bin: PgBin):
+def test_twophase(zenith_cli,
+                  pageserver: ZenithPageserver,
+                  postgres: PostgresFactory,
+                  pg_bin: PgBin):
     zenith_cli.run(["branch", "test_twophase", "empty"])
 
     pg = postgres.create_start('test_twophase', config_lines=['max_prepared_transactions=5'])
@@ -79,8 +82,8 @@ def test_twophase(zenith_cli, pageserver: ZenithPageserver, postgres: PostgresFa
     cur2.execute("ROLLBACK PREPARED 'insert_two'")
 
     cur2.execute('SELECT * FROM foo')
-    assert cur2.fetchall() == [('one',), ('three',)]
+    assert cur2.fetchall() == [('one', ), ('three', )]
 
     # Only one committed insert is visible on the original branch
     cur.execute('SELECT * FROM foo')
-    assert cur.fetchall() == [('three',)]
+    assert cur.fetchall() == [('three', )]

test_runner/batch_others/test_vm_bits.py

@@ -3,11 +3,16 @@ from fixtures.log_helper import log
 
 pytest_plugins = ("fixtures.zenith_fixtures")
 
+
 #
 # Test that the VM bit is cleared correctly at a HEAP_DELETE and
 # HEAP_UPDATE record.
 #
-def test_vm_bit_clear(pageserver: ZenithPageserver, postgres: PostgresFactory, pg_bin, zenith_cli, base_dir):
+def test_vm_bit_clear(pageserver: ZenithPageserver,
+                      postgres: PostgresFactory,
+                      pg_bin,
+                      zenith_cli,
+                      base_dir):
     # Create a branch for us
     zenith_cli.run(["branch", "test_vm_bit_clear", "empty"])
     pg = postgres.create_start('test_vm_bit_clear')
@@ -49,13 +54,12 @@ def test_vm_bit_clear(pageserver: ZenithPageserver, postgres: PostgresFactory, p
     ''')
 
     cur.execute('SELECT * FROM vmtest_delete WHERE id = 1')
-    assert(cur.fetchall() == []);
+    assert (cur.fetchall() == [])
     cur.execute('SELECT * FROM vmtest_update WHERE id = 1')
-    assert(cur.fetchall() == []);
+    assert (cur.fetchall() == [])
 
     cur.close()
 
-
     # Check the same thing on the branch that we created right after the DELETE
     #
     # As of this writing, the code in smgrwrite() creates a full-page image whenever
@@ -75,6 +79,6 @@ def test_vm_bit_clear(pageserver: ZenithPageserver, postgres: PostgresFactory, p
     ''')
 
     cur_new.execute('SELECT * FROM vmtest_delete WHERE id = 1')
-    assert(cur_new.fetchall() == []);
+    assert (cur_new.fetchall() == [])
     cur_new.execute('SELECT * FROM vmtest_update WHERE id = 1')
-    assert(cur_new.fetchall() == []);
+    assert (cur_new.fetchall() == [])

test_runner/batch_others/test_wal_acceptor.py

@@ -16,7 +16,10 @@ pytest_plugins = ("fixtures.zenith_fixtures")
 
 # basic test, write something in setup with wal acceptors, ensure that commits
 # succeed and data is written
-def test_normal_work(zenith_cli, pageserver: ZenithPageserver, postgres: PostgresFactory, wa_factory):
+def test_normal_work(zenith_cli,
+                     pageserver: ZenithPageserver,
+                     postgres: PostgresFactory,
+                     wa_factory):
     zenith_cli.run(["branch", "test_wal_acceptors_normal_work", "empty"])
     wa_factory.start_n_new(3)
     pg = postgres.create_start('test_wal_acceptors_normal_work',
@@ -34,7 +37,10 @@ def test_normal_work(zenith_cli, pageserver: ZenithPageserver, postgres: Postgre
 
 # Run page server and multiple acceptors, and multiple compute nodes running
 # against different timelines.
-def test_many_timelines(zenith_cli, pageserver: ZenithPageserver, postgres: PostgresFactory, wa_factory):
+def test_many_timelines(zenith_cli,
+                        pageserver: ZenithPageserver,
+                        postgres: PostgresFactory,
+                        wa_factory):
     n_timelines = 2
 
     wa_factory.start_n_new(3)
@@ -66,7 +72,10 @@ def test_many_timelines(zenith_cli, pageserver: ZenithPageserver, postgres: Post
 # Check that dead minority doesn't prevent the commits: execute insert n_inserts
 # times, with fault_probability chance of getting a wal acceptor down or up
 # along the way. 2 of 3 are always alive, so the work keeps going.
-def test_restarts(zenith_cli, pageserver: ZenithPageserver, postgres: PostgresFactory, wa_factory: WalAcceptorFactory):
+def test_restarts(zenith_cli,
+                  pageserver: ZenithPageserver,
+                  postgres: PostgresFactory,
+                  wa_factory: WalAcceptorFactory):
     fault_probability = 0.01
     n_inserts = 1000
     n_acceptors = 3
@@ -177,7 +186,11 @@ def stop_value():
 
 
 # do inserts while concurrently getting up/down subsets of acceptors
-def test_race_conditions(zenith_cli, pageserver: ZenithPageserver, postgres: PostgresFactory, wa_factory, stop_value):
+def test_race_conditions(zenith_cli,
+                         pageserver: ZenithPageserver,
+                         postgres: PostgresFactory,
+                         wa_factory,
+                         stop_value):
 
     wa_factory.start_n_new(3)
 
@@ -318,6 +331,5 @@ def test_timeline_status(zenith_cli, pageserver, postgres, wa_factory: WalAccept
     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
+    epoch_after_reboot = wa_http_cli.timeline_status(tenant_id, timeline_id).acceptor_epoch
     assert epoch_after_reboot > epoch

test_runner/batch_others/test_wal_acceptor_async.py

@@ -19,13 +19,16 @@ class BankClient(object):
     async def initdb(self):
         await self.conn.execute('DROP TABLE IF EXISTS bank_accs')
         await self.conn.execute('CREATE TABLE bank_accs(uid int primary key, amount int)')
-        await self.conn.execute('''
+        await self.conn.execute(
+            '''
             INSERT INTO bank_accs
             SELECT *, $1 FROM generate_series(0, $2)
-        ''', self.init_amount, self.n_accounts - 1)
+        ''',
+            self.init_amount,
+            self.n_accounts - 1)
         await self.conn.execute('DROP TABLE IF EXISTS bank_log')
         await self.conn.execute('CREATE TABLE bank_log(from_uid int, to_uid int, amount int)')
-        
+
         # TODO: Remove when https://github.com/zenithdb/zenith/issues/644 is fixed
         await self.conn.execute('ALTER TABLE bank_accs SET (autovacuum_enabled = false)')
         await self.conn.execute('ALTER TABLE bank_log SET (autovacuum_enabled = false)')
@@ -34,6 +37,7 @@ class BankClient(object):
         row = await self.conn.fetchrow('SELECT sum(amount) AS sum FROM bank_accs')
         assert row['sum'] == self.n_accounts * self.init_amount
 
+
 async def bank_transfer(conn: asyncpg.Connection, from_uid, to_uid, amount):
     # avoid deadlocks by sorting uids
     if from_uid > to_uid:
@@ -42,16 +46,22 @@ async def bank_transfer(conn: asyncpg.Connection, from_uid, to_uid, amount):
     async with conn.transaction():
         await conn.execute(
             'UPDATE bank_accs SET amount = amount + ($1) WHERE uid = $2',
-            amount, to_uid,
+            amount,
+            to_uid,
         )
         await conn.execute(
             'UPDATE bank_accs SET amount = amount - ($1) WHERE uid = $2',
-            amount, from_uid,
+            amount,
+            from_uid,
         )
-        await conn.execute('INSERT INTO bank_log VALUES ($1, $2, $3)',
-            from_uid, to_uid, amount,
+        await conn.execute(
+            'INSERT INTO bank_log VALUES ($1, $2, $3)',
+            from_uid,
+            to_uid,
+            amount,
         )
 
+
 class WorkerStats(object):
     def __init__(self, n_workers):
         self.counters = [0] * n_workers
@@ -114,7 +124,6 @@ async def run_restarts_under_load(pg: Postgres, acceptors: List[WalAcceptor], n_
         worker = run_random_worker(stats, pg, worker_id, bank.n_accounts, max_transfer)
         workers.append(asyncio.create_task(worker))
 
-
     for it in range(iterations):
         victim = acceptors[it % len(acceptors)]
         victim.stop()
@@ -122,10 +131,7 @@ async def run_restarts_under_load(pg: Postgres, acceptors: List[WalAcceptor], n_
         # Wait till previous victim recovers so it is ready for the next
         # iteration by making any writing xact.
         conn = await pg.connect_async()
-        await conn.execute(
-            'UPDATE bank_accs SET amount = amount WHERE uid = 1',
-            timeout=120
-        )
+        await conn.execute('UPDATE bank_accs SET amount = amount WHERE uid = 1', timeout=120)
         await conn.close()
 
         stats.reset()
@@ -145,7 +151,9 @@ async def run_restarts_under_load(pg: Postgres, acceptors: List[WalAcceptor], n_
 
 
 # restart acceptors one by one, while executing and validating bank transactions
-def test_restarts_under_load(zenith_cli, pageserver: ZenithPageserver, postgres: PostgresFactory,
+def test_restarts_under_load(zenith_cli,
+                             pageserver: ZenithPageserver,
+                             postgres: PostgresFactory,
                              wa_factory: WalAcceptorFactory):
 
     wa_factory.start_n_new(3)

test_runner/batch_others/test_zenith_cli.py

@@ -23,8 +23,11 @@ def helper_compare_branch_list(page_server_cur, zenith_cli, initial_tenant: str)
 
     res = zenith_cli.run(["branch", f"--tenantid={initial_tenant}"])
     res.check_returncode()
-    branches_cli_with_tenant_arg = sorted(map(lambda b: b.split(':')[-1].strip(), res.stdout.strip().split("\n")))
-    branches_cli_with_tenant_arg = [b for b in branches_cli if b.startswith('test_cli_') or b in ('empty', 'main')]
+    branches_cli_with_tenant_arg = sorted(
+        map(lambda b: b.split(':')[-1].strip(), res.stdout.strip().split("\n")))
+    branches_cli_with_tenant_arg = [
+        b for b in branches_cli if b.startswith('test_cli_') or b in ('empty', 'main')
+    ]
 
     assert branches_api == branches_cli == branches_cli_with_tenant_arg
 
@@ -54,6 +57,7 @@ def test_cli_branch_list(pageserver: ZenithPageserver, zenith_cli):
     assert 'test_cli_branch_list_main' in branches_cli
     assert 'test_cli_branch_list_nested' in branches_cli
 
+
 def helper_compare_tenant_list(page_server_cur, zenith_cli: ZenithCli):
     page_server_cur.execute(f'tenant_list')
     tenants_api = sorted(json.loads(page_server_cur.fetchone()[0]))

test_runner/batch_pg_regress/test_isolation.py

@@ -6,8 +6,14 @@ from fixtures.zenith_fixtures import ZenithPageserver, PostgresFactory
 pytest_plugins = ("fixtures.zenith_fixtures")
 
 
-def test_isolation(pageserver: ZenithPageserver, postgres: PostgresFactory, pg_bin, zenith_cli, test_output_dir, pg_distrib_dir,
-                   base_dir, capsys):
+def test_isolation(pageserver: ZenithPageserver,
+                   postgres: PostgresFactory,
+                   pg_bin,
+                   zenith_cli,
+                   test_output_dir,
+                   pg_distrib_dir,
+                   base_dir,
+                   capsys):
 
     # Create a branch for us
     zenith_cli.run(["branch", "test_isolation", "empty"])

test_runner/batch_pg_regress/test_pg_regress.py

@@ -6,8 +6,14 @@ from fixtures.zenith_fixtures import PostgresFactory, ZenithPageserver, check_re
 pytest_plugins = ("fixtures.zenith_fixtures")
 
 
-def test_pg_regress(pageserver: ZenithPageserver, postgres: PostgresFactory, pg_bin, zenith_cli, test_output_dir, pg_distrib_dir,
-                    base_dir, capsys):
+def test_pg_regress(pageserver: ZenithPageserver,
+                    postgres: PostgresFactory,
+                    pg_bin,
+                    zenith_cli,
+                    test_output_dir,
+                    pg_distrib_dir,
+                    base_dir,
+                    capsys):
 
     # Create a branch for us
     zenith_cli.run(["branch", "test_pg_regress", "empty"])

test_runner/batch_pg_regress/test_zenith_regress.py

@@ -7,8 +7,14 @@ from fixtures.log_helper import log
 pytest_plugins = ("fixtures.zenith_fixtures")
 
 
-def test_zenith_regress(postgres: PostgresFactory, pg_bin, zenith_cli, test_output_dir, pg_distrib_dir,
-                        base_dir, capsys, pageserver_port: PageserverPort):
+def test_zenith_regress(postgres: PostgresFactory,
+                        pg_bin,
+                        zenith_cli,
+                        test_output_dir,
+                        pg_distrib_dir,
+                        base_dir,
+                        capsys,
+                        pageserver_port: PageserverPort):
 
     # Create a branch for us
     zenith_cli.run(["branch", "test_zenith_regress", "empty"])

test_runner/fixtures/benchmark_fixture.py

@@ -24,7 +24,6 @@ from typing import Any, Callable, Dict, Iterator, List, Optional, TypeVar, cast
 from typing_extensions import Literal
 
 from .utils import (get_self_dir, mkdir_if_needed, subprocess_capture)
-
 """
 This file contains fixtures for micro-benchmarks.
 
@@ -55,7 +54,6 @@ in the test initialization, or measure disk usage after the test query.
 
 """
 
-
 # All the results are collected in this list, as a tuple:
 # (test_name: str, metric_name: str, metric_value: float, unit: str)
 #
@@ -65,6 +63,7 @@ in the test initialization, or measure disk usage after the test query.
 global zenbenchmark_results
 zenbenchmark_results = []
 
+
 class ZenithBenchmarkResults:
     """ An object for recording benchmark results. """
     def __init__(self):
@@ -77,6 +76,7 @@ class ZenithBenchmarkResults:
 
         self.results.append((test_name, metric_name, metric_value, unit))
 
+
 # Session scope fixture that initializes the results object
 @pytest.fixture(autouse=True, scope='session')
 def zenbenchmark_global(request) -> Iterator[ZenithBenchmarkResults]:
@@ -88,6 +88,7 @@ def zenbenchmark_global(request) -> Iterator[ZenithBenchmarkResults]:
 
     yield zenbenchmark_results
 
+
 class ZenithBenchmarker:
     """
     An object for recording benchmark results. This is created for each test
@@ -103,7 +104,6 @@ class ZenithBenchmarker:
         """
         self.results.record(self.request.node.name, metric_name, metric_value, unit)
 
-
     @contextmanager
     def record_duration(self, metric_name):
         """
@@ -134,7 +134,8 @@ class ZenithBenchmarker:
         # The metric should be an integer, as it's a number of bytes. But in general
         # all prometheus metrics are floats. So to be pedantic, read it as a float
         # and round to integer.
-        matches = re.search(r'^pageserver_disk_io_bytes{io_operation="write"} (\S+)$', all_metrics,
+        matches = re.search(r'^pageserver_disk_io_bytes{io_operation="write"} (\S+)$',
+                            all_metrics,
                             re.MULTILINE)
         return int(round(float(matches.group(1))))
 
@@ -145,8 +146,7 @@ class ZenithBenchmarker:
         # Fetch all the exposed prometheus metrics from page server
         all_metrics = pageserver.http_client().get_metrics()
         # See comment in get_io_writes()
-        matches = re.search(r'^pageserver_maxrss_kb (\S+)$', all_metrics,
-                            re.MULTILINE)
+        matches = re.search(r'^pageserver_maxrss_kb (\S+)$', all_metrics, re.MULTILINE)
         return int(round(float(matches.group(1))))
 
     def get_timeline_size(self, repo_dir: str, tenantid: str, timelineid: str):
@@ -171,7 +171,11 @@ class ZenithBenchmarker:
         yield
         after = self.get_io_writes(pageserver)
 
-        self.results.record(self.request.node.name, metric_name, round((after - before) / (1024 * 1024)), 'MB')
+        self.results.record(self.request.node.name,
+                            metric_name,
+                            round((after - before) / (1024 * 1024)),
+                            'MB')
+
 
 @pytest.fixture(scope='function')
 def zenbenchmark(zenbenchmark_global, request) -> Iterator[ZenithBenchmarker]:
@@ -185,9 +189,7 @@ def zenbenchmark(zenbenchmark_global, request) -> Iterator[ZenithBenchmarker]:
 
 # Hook to print the results at the end
 @pytest.hookimpl(hookwrapper=True)
-def pytest_terminal_summary(
-    terminalreporter: TerminalReporter, exitstatus: int, config: Config
-):
+def pytest_terminal_summary(terminalreporter: TerminalReporter, exitstatus: int, config: Config):
     yield
 
     global zenbenchmark_results

test_runner/fixtures/log_helper.py

@@ -1,6 +1,5 @@
 import logging
 import logging.config
-
 """
 This file configures logging to use in python tests.
 Logs are automatically captured and shown in their
@@ -27,17 +26,19 @@ LOGGING = {
             "level": "INFO"
         },
         "root.wal_acceptor_async": {
-            "level": "INFO" # a lot of logs on DEBUG level
+            "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()
 

test_runner/fixtures/utils.py

@@ -4,6 +4,7 @@ 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. """
     return os.path.dirname(os.path.abspath(__file__))
@@ -58,6 +59,7 @@ def global_counter() -> int:
     _global_counter += 1
     return _global_counter
 
+
 def lsn_to_hex(num: int) -> str:
     """ Convert lsn from int to standard hex notation. """
     return "{:X}/{:X}".format(num >> 32, num & 0xffffffff)

test_runner/fixtures/zenith_fixtures.py

@@ -1,5 +1,5 @@
 from dataclasses import dataclass
-from functools import cached_property
+from cached_property import cached_property
 import asyncpg
 import os
 import pathlib
@@ -13,9 +13,8 @@ import signal
 import subprocess
 import time
 import filecmp
-import difflib
 
-from contextlib import closing
+from contextlib import closing, suppress
 from pathlib import Path
 from dataclasses import dataclass
 
@@ -28,7 +27,6 @@ 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.
@@ -56,14 +54,15 @@ DEFAULT_POSTGRES_DIR = 'tmp_install'
 BASE_PORT = 15000
 WORKER_PORT_NUM = 100
 
+
 def pytest_configure(config):
     """
     Ensure that no unwanted daemons are running before we start testing.
     Check that we do not owerflow available ports range.
     """
     numprocesses = config.getoption('numprocesses')
-    if numprocesses is not None and BASE_PORT + numprocesses * WORKER_PORT_NUM > 32768: # do not use ephemeral ports
-         raise Exception('Too many workers configured. Cannot distrubute ports for services.')
+    if numprocesses is not None and BASE_PORT + numprocesses * WORKER_PORT_NUM > 32768:  # do not use ephemeral ports
+        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']
@@ -107,7 +106,11 @@ class PgProtocol:
         self.port = port
         self.username = username or "zenith_admin"
 
-    def connstr(self, *, dbname: str = 'postgres', username: Optional[str] = None, password: Optional[str] = None) -> str:
+    def connstr(self,
+                *,
+                dbname: str = 'postgres',
+                username: Optional[str] = None,
+                password: Optional[str] = None) -> str:
         """
         Build a libpq connection string for the Postgres instance.
         """
@@ -119,7 +122,12 @@ class PgProtocol:
         return f'{res} password={password}'
 
     # autocommit=True here by default because that's what we need most of the time
-    def connect(self, *, autocommit=True, dbname: str = 'postgres', username: Optional[str] = None, password: Optional[str] = None) -> PgConnection:
+    def connect(self,
+                *,
+                autocommit=True,
+                dbname: str = 'postgres',
+                username: Optional[str] = None,
+                password: Optional[str] = None) -> PgConnection:
         """
         Connect to the node.
         Returns psycopg2's connection object.
@@ -135,7 +143,11 @@ class PgProtocol:
         conn.autocommit = autocommit
         return conn
 
-    async def connect_async(self, *, dbname: str = 'postgres', username: Optional[str] = None, password: Optional[str] = None) -> asyncpg.Connection:
+    async def connect_async(self,
+                            *,
+                            dbname: str = 'postgres',
+                            username: Optional[str] = None,
+                            password: Optional[str] = None) -> asyncpg.Connection:
         """
         Connect to the node from async python.
         Returns asyncpg's connection object.
@@ -201,11 +213,11 @@ class ZenithCli:
         # Interceipt CalledProcessError and print more info
         try:
             res = subprocess.run(args,
-                                env=self.env,
-                                check=True,
-                                universal_newlines=True,
-                                stdout=subprocess.PIPE,
-                                stderr=subprocess.PIPE)
+                                 env=self.env,
+                                 check=True,
+                                 universal_newlines=True,
+                                 stdout=subprocess.PIPE,
+                                 stderr=subprocess.PIPE)
         except subprocess.CalledProcessError as exc:
             # this way command output will be in recorded and shown in CI in failure message
             msg = f"""\
@@ -235,40 +247,36 @@ 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",
-            json={
-                'tenant_id': tenant_id.hex,
-                'name': name,
-                'start_point': start_point,
-            }
-        )
+        res = self.post(f"http://localhost:{self.port}/v1/branch",
+                        json={
+                            'tenant_id': tenant_id.hex,
+                            'name': name,
+                            'start_point': start_point,
+                        })
         res.raise_for_status()
         return res.json()
 
     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}",
-        )
+        res = self.get(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 +285,7 @@ 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
 
@@ -299,7 +307,11 @@ class AuthKeys:
         return token
 
     def generate_tenant_token(self, tenant_id):
-        token = jwt.encode({"scope": "tenant", "tenant_id": tenant_id}, self.priv, algorithm="RS256")
+        token = jwt.encode({
+            "scope": "tenant", "tenant_id": tenant_id
+        },
+                           self.priv,
+                           algorithm="RS256")
 
         if isinstance(token, bytes):
             token = token.decode()
@@ -324,6 +336,7 @@ def worker_base_port(worker_seq_no: int):
     # so workers have disjoint set of ports for services
     return BASE_PORT + worker_seq_no * WORKER_PORT_NUM
 
+
 class PortDistributor:
     def __init__(self, base_port: int, port_number: int) -> None:
         self.iterator = iter(range(base_port, base_port + port_number))
@@ -332,13 +345,15 @@ class PortDistributor:
         try:
             return next(self.iterator)
         except StopIteration:
-            raise RuntimeError('port range configured for test is exhausted, consider enlarging the range')
+            raise RuntimeError(
+                'port range configured for test is exhausted, consider enlarging the range')
 
 
 @zenfixture
 def port_distributor(worker_base_port):
     return PortDistributor(base_port=worker_base_port, port_number=WORKER_PORT_NUM)
 
+
 @dataclass
 class PageserverPort:
     pg: int
@@ -348,19 +363,23 @@ 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
         self.repo_dir = repo_dir
-        self.service_port = port # do not shadow PgProtocol.port which is just int
+        self.service_port = port  # do not shadow PgProtocol.port which is just int
 
     def init(self, enable_auth: bool = False) -> 'ZenithPageserver':
         """
         Initialize the repository, i.e. run "zenith init".
         Returns self.
         """
-        cmd = ['init', f'--pageserver-pg-port={self.service_port.pg}', f'--pageserver-http-port={self.service_port.http}']
+        cmd = [
+            'init',
+            f'--pageserver-pg-port={self.service_port.pg}',
+            f'--pageserver-http-port={self.service_port.http}'
+        ]
         if enable_auth:
             cmd.append('--enable-auth')
         self.zenith_cli.run(cmd)
@@ -420,8 +439,6 @@ class ZenithPageserver(PgProtocol):
         )
 
 
-
-
 @zenfixture
 def pageserver_port(port_distributor: PortDistributor) -> PageserverPort:
     pg = port_distributor.get_port()
@@ -431,7 +448,8 @@ def pageserver_port(port_distributor: PortDistributor) -> PageserverPort:
 
 
 @zenfixture
-def pageserver(zenith_cli: ZenithCli, repo_dir: str, pageserver_port: PageserverPort) -> Iterator[ZenithPageserver]:
+def pageserver(zenith_cli: ZenithCli, repo_dir: str,
+               pageserver_port: PageserverPort) -> Iterator[ZenithPageserver]:
     """
     The 'pageserver' fixture provides a Page Server that's up and running.
 
@@ -443,7 +461,8 @@ def pageserver(zenith_cli: ZenithCli, repo_dir: str, pageserver_port: Pageserver
     By convention, the test branches are named after the tests. For example,
     test called 'test_foo' would create and use branches with the 'test_foo' prefix.
     """
-    ps = ZenithPageserver(zenith_cli=zenith_cli, repo_dir=repo_dir, port=pageserver_port).init().start()
+    ps = ZenithPageserver(zenith_cli=zenith_cli, repo_dir=repo_dir,
+                          port=pageserver_port).init().start()
     # For convenience in tests, create a branch from the freshly-initialized cluster.
     zenith_cli.run(["branch", "empty", "main"])
 
@@ -453,6 +472,7 @@ def pageserver(zenith_cli: ZenithCli, repo_dir: str, pageserver_port: Pageserver
     log.info('Starting pageserver cleanup')
     ps.stop(True)
 
+
 class PgBin:
     """ A helper class for executing postgres binaries """
     def __init__(self, log_dir: str, pg_distrib_dir: str):
@@ -514,9 +534,11 @@ class PgBin:
 def pg_bin(test_output_dir: str, pg_distrib_dir: str) -> PgBin:
     return PgBin(test_output_dir, pg_distrib_dir)
 
+
 @pytest.fixture
 def pageserver_auth_enabled(zenith_cli: ZenithCli, repo_dir: str, pageserver_port: PageserverPort):
-    with ZenithPageserver(zenith_cli=zenith_cli, repo_dir=repo_dir, port=pageserver_port).init(enable_auth=True).start() as ps:
+    with ZenithPageserver(zenith_cli=zenith_cli, repo_dir=repo_dir,
+                          port=pageserver_port).init(enable_auth=True).start() as ps:
         # For convenience in tests, create a branch from the freshly-initialized cluster.
         zenith_cli.run(["branch", "empty", "main"])
         yield ps
@@ -524,21 +546,27 @@ 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)
+    def __init__(self,
+                 zenith_cli: ZenithCli,
+                 repo_dir: str,
+                 pg_bin: PgBin,
+                 tenant_id: str,
+                 port: int):
+        super().__init__(host='localhost', port=port)
 
         self.zenith_cli = zenith_cli
         self.running = False
         self.repo_dir = repo_dir
-        self.branch: Optional[str] = None  # dubious, see asserts below
-        self.pgdata_dir: Optional[str] = None # Path to computenode PGDATA
+        self.node_name: Optional[str] = None  # dubious, see asserts below
+        self.pgdata_dir: Optional[str] = None  # Path to computenode PGDATA
         self.tenant_id = tenant_id
         self.pg_bin = pg_bin
-        # path to conf is <repo_dir>/pgdatadirs/tenants/<tenant_id>/<branch_name>/postgresql.conf
+        # path to conf is <repo_dir>/pgdatadirs/tenants/<tenant_id>/<node_name>/postgresql.conf
 
     def create(
         self,
-        branch: str,
+        node_name: str,
+        branch: Optional[str] = None,
         wal_acceptors: Optional[str] = None,
         config_lines: Optional[List[str]] = None,
     ) -> 'Postgres':
@@ -552,9 +580,19 @@ class Postgres(PgProtocol):
         if not config_lines:
             config_lines = []
 
-        self.zenith_cli.run(['pg', 'create', branch, f'--tenantid={self.tenant_id}', f'--port={self.port}'])
-        self.branch = branch
-        path = pathlib.Path('pgdatadirs') / 'tenants' / self.tenant_id / self.branch
+        if branch is None:
+            branch = node_name
+
+        self.zenith_cli.run([
+            'pg',
+            'create',
+            f'--tenantid={self.tenant_id}',
+            f'--port={self.port}',
+            node_name,
+            branch
+        ])
+        self.node_name = node_name
+        path = pathlib.Path('pgdatadirs') / 'tenants' / self.tenant_id / self.node_name
         self.pgdata_dir = os.path.join(self.repo_dir, path)
 
         if wal_acceptors is not None:
@@ -571,11 +609,12 @@ class Postgres(PgProtocol):
         Returns self.
         """
 
-        assert self.branch is not None
+        assert self.node_name is not None
 
-        log.info(f"Starting postgres on branch {self.branch}")
+        log.info(f"Starting postgres node {self.node_name}")
 
-        run_result = self.zenith_cli.run(['pg', 'start', self.branch, f'--tenantid={self.tenant_id}', f'--port={self.port}'])
+        run_result = self.zenith_cli.run(
+            ['pg', 'start', f'--tenantid={self.tenant_id}', f'--port={self.port}', self.node_name])
         self.running = True
 
         log.info(f"stdout: {run_result.stdout}")
@@ -584,7 +623,7 @@ class Postgres(PgProtocol):
 
     def pg_data_dir_path(self) -> str:
         """ Path to data directory """
-        path = pathlib.Path('pgdatadirs') / 'tenants' / self.tenant_id / self.branch
+        path = pathlib.Path('pgdatadirs') / 'tenants' / self.tenant_id / self.node_name
         return os.path.join(self.repo_dir, path)
 
     def pg_xact_dir_path(self) -> str:
@@ -641,8 +680,8 @@ class Postgres(PgProtocol):
         """
 
         if self.running:
-            assert self.branch is not None
-            self.zenith_cli.run(['pg', 'stop', self.branch, f'--tenantid={self.tenant_id}'])
+            assert self.node_name is not None
+            self.zenith_cli.run(['pg', 'stop', self.node_name, f'--tenantid={self.tenant_id}'])
             self.running = False
 
         return self
@@ -653,15 +692,17 @@ class Postgres(PgProtocol):
         Returns self.
         """
 
-        assert self.branch is not None
+        assert self.node_name is not None
         assert self.tenant_id is not None
-        self.zenith_cli.run(['pg', 'stop', '--destroy', self.branch, f'--tenantid={self.tenant_id}'])
+        self.zenith_cli.run(
+            ['pg', 'stop', '--destroy', self.node_name, f'--tenantid={self.tenant_id}'])
 
         return self
 
     def create_start(
         self,
-        branch: str,
+        node_name: str,
+        branch: Optional[str] = None,
         wal_acceptors: Optional[str] = None,
         config_lines: Optional[List[str]] = None,
     ) -> 'Postgres':
@@ -672,6 +713,7 @@ class Postgres(PgProtocol):
         """
 
         self.create(
+            node_name=node_name,
             branch=branch,
             wal_acceptors=wal_acceptors,
             config_lines=config_lines,
@@ -685,9 +727,15 @@ class Postgres(PgProtocol):
     def __exit__(self, exc_type, exc, tb):
         self.stop()
 
+
 class PostgresFactory:
     """ An object representing multiple running postgres daemons. """
-    def __init__(self, zenith_cli: ZenithCli, repo_dir: str, pg_bin: PgBin, initial_tenant: str, port_distributor: PortDistributor):
+    def __init__(self,
+                 zenith_cli: ZenithCli,
+                 repo_dir: str,
+                 pg_bin: PgBin,
+                 initial_tenant: str,
+                 port_distributor: PortDistributor):
         self.zenith_cli = zenith_cli
         self.repo_dir = repo_dir
         self.num_instances = 0
@@ -696,13 +744,13 @@ class PostgresFactory:
         self.port_distributor = port_distributor
         self.pg_bin = pg_bin
 
-    def create_start(
-        self,
-        branch: str = "main",
-        tenant_id: Optional[str] = None,
-        wal_acceptors: Optional[str] = None,
-        config_lines: Optional[List[str]] = None
-    ) -> Postgres:
+    def create_start(self,
+                     node_name: str = "main",
+                     branch: Optional[str] = None,
+                     tenant_id: Optional[str] = None,
+                     wal_acceptors: Optional[str] = None,
+                     config_lines: Optional[List[str]] = None) -> Postgres:
+
         pg = Postgres(
             zenith_cli=self.zenith_cli,
             repo_dir=self.repo_dir,
@@ -714,18 +762,18 @@ class PostgresFactory:
         self.instances.append(pg)
 
         return pg.create_start(
+            node_name=node_name,
             branch=branch,
             wal_acceptors=wal_acceptors,
             config_lines=config_lines,
         )
 
-    def create(
-        self,
-        branch: str = "main",
-        tenant_id: Optional[str] = None,
-        wal_acceptors: Optional[str] = None,
-        config_lines: Optional[List[str]] = None
-    ) -> Postgres:
+    def create(self,
+               node_name: str = "main",
+               branch: Optional[str] = None,
+               tenant_id: Optional[str] = None,
+               wal_acceptors: Optional[str] = None,
+               config_lines: Optional[List[str]] = None) -> Postgres:
 
         pg = Postgres(
             zenith_cli=self.zenith_cli,
@@ -739,18 +787,17 @@ class PostgresFactory:
         self.instances.append(pg)
 
         return pg.create(
+            node_name=node_name,
             branch=branch,
             wal_acceptors=wal_acceptors,
             config_lines=config_lines,
         )
 
-    def config(
-        self,
-        branch: str = "main",
-        tenant_id: Optional[str] = None,
-        wal_acceptors: Optional[str] = None,
-        config_lines: Optional[List[str]] = None
-    ) -> Postgres:
+    def config(self,
+               node_name: str = "main",
+               tenant_id: Optional[str] = None,
+               wal_acceptors: Optional[str] = None,
+               config_lines: Optional[List[str]] = None) -> Postgres:
 
         pg = Postgres(
             zenith_cli=self.zenith_cli,
@@ -764,7 +811,7 @@ class PostgresFactory:
         self.instances.append(pg)
 
         return pg.config(
-            branch=branch,
+            node_name=node_name,
             wal_acceptors=wal_acceptors,
             config_lines=config_lines,
         )
@@ -775,13 +822,18 @@ class PostgresFactory:
 
         return self
 
+
 @zenfixture
 def initial_tenant(pageserver: ZenithPageserver):
     return pageserver.initial_tenant
 
 
 @zenfixture
-def postgres(zenith_cli: ZenithCli, initial_tenant: str, repo_dir: str, pg_bin: PgBin, port_distributor: PortDistributor) -> Iterator[PostgresFactory]:
+def postgres(zenith_cli: ZenithCli,
+             initial_tenant: str,
+             repo_dir: str,
+             pg_bin: PgBin,
+             port_distributor: PortDistributor) -> Iterator[PostgresFactory]:
     pgfactory = PostgresFactory(
         zenith_cli=zenith_cli,
         repo_dir=repo_dir,
@@ -796,6 +848,7 @@ def postgres(zenith_cli: ZenithCli, initial_tenant: str, repo_dir: str, pg_bin:
     log.info('Starting postgres cleanup')
     pgfactory.stop_all()
 
+
 def read_pid(path: Path):
     """ Read content of file into number """
     return int(path.read_text())
@@ -806,29 +859,31 @@ class WalAcceptorPort:
     pg: int
     http: int
 
+
 @dataclass
 class WalAcceptor:
     """ An object representing a running wal acceptor daemon. """
     wa_bin_path: Path
     data_dir: Path
     port: WalAcceptorPort
-    num: int # identifier for logging
+    num: int  # identifier for logging
     pageserver_port: int
     auth_token: Optional[str] = None
 
     def start(self) -> 'WalAcceptor':
         # create data directory if not exists
         self.data_dir.mkdir(parents=True, exist_ok=True)
-        self.pidfile.unlink(missing_ok=True)
+        with suppress(FileNotFoundError):
+            self.pidfile.unlink()
 
         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)))
         env = {'PAGESERVER_AUTH_TOKEN': self.auth_token} if self.auth_token else None
@@ -876,10 +931,11 @@ class WalAcceptor:
             os.kill(pid, signal.SIGTERM)
         except Exception:
             # TODO: cleanup pid file on exit in wal acceptor
-            pass # pidfile might be obsolete
+            pass  # pidfile might be obsolete
         return self
 
-    def append_logical_message(self, tenant_id: str, timeline_id: str, request: Dict[str, Any]) -> Dict[str, Any]:
+    def append_logical_message(self, tenant_id: str, timeline_id: str,
+                               request: Dict[str, Any]) -> Dict[str, Any]:
         """
         Send JSON_CTRL query to append LogicalMessage to WAL and modify 
         safekeeper state. It will construct LogicalMessage from provided
@@ -888,7 +944,7 @@ 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
@@ -907,7 +963,11 @@ class WalAcceptor:
 
 class WalAcceptorFactory:
     """ An object representing multiple running wal acceptors. """
-    def __init__(self, zenith_binpath: Path, data_dir: Path, pageserver_port: int, port_distributor: PortDistributor):
+    def __init__(self,
+                 zenith_binpath: Path,
+                 data_dir: Path,
+                 pageserver_port: int,
+                 port_distributor: PortDistributor):
         self.wa_bin_path = zenith_binpath / 'safekeeper'
         self.data_dir = data_dir
         self.instances: List[WalAcceptor] = []
@@ -949,11 +1009,14 @@ 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
-def wa_factory(zenith_binpath: str, repo_dir: str, pageserver_port: PageserverPort, port_distributor: PortDistributor) -> Iterator[WalAcceptorFactory]:
+def wa_factory(zenith_binpath: str,
+               repo_dir: str,
+               pageserver_port: PageserverPort,
+               port_distributor: PortDistributor) -> Iterator[WalAcceptorFactory]:
     """ Gives WalAcceptorFactory providing wal acceptors. """
     wafactory = WalAcceptorFactory(
         zenith_binpath=Path(zenith_binpath),
@@ -966,20 +1029,22 @@ def wa_factory(zenith_binpath: str, repo_dir: str, pageserver_port: PageserverPo
     log.info('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()
+        self.get(f"http://localhost:{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 = self.get(f"http://localhost:{self.port}/v1/timeline/{tenant_id}/{timeline_id}")
         res.raise_for_status()
         resj = res.json()
         return PageserverTimelineStatus(acceptor_epoch=resj['acceptor_state']['epoch'])
@@ -1083,6 +1148,7 @@ class TenantFactory:
 def tenant_factory(zenith_cli: ZenithCli):
     return TenantFactory(zenith_cli)
 
+
 #
 # Test helpers
 #
@@ -1093,8 +1159,15 @@ def list_files_to_compare(pgdata_dir: str):
             rel_dir = os.path.relpath(root, pgdata_dir)
             # Skip some dirs and files we don't want to compare
             skip_dirs = ['pg_wal', 'pg_stat', 'pg_stat_tmp', 'pg_subtrans', 'pg_logical']
-            skip_files = ['pg_internal.init', 'pg.log', 'zenith.signal', 'postgresql.conf',
-                        'postmaster.opts', 'postmaster.pid', 'pg_control']
+            skip_files = [
+                'pg_internal.init',
+                'pg.log',
+                'zenith.signal',
+                'postgresql.conf',
+                'postmaster.opts',
+                'postmaster.pid',
+                'pg_control'
+            ]
             if rel_dir not in skip_dirs and filename not in skip_files:
                 rel_file = os.path.join(rel_dir, filename)
                 pgdata_files.append(rel_file)
@@ -1103,8 +1176,12 @@ def list_files_to_compare(pgdata_dir: str):
     log.info(pgdata_files)
     return pgdata_files
 
+
 # pg is the existing and running compute node, that we want to compare with a basebackup
-def check_restored_datadir_content(zenith_cli: ZenithCli, test_output_dir: str, pg: Postgres, pageserver_pg_port: int):
+def check_restored_datadir_content(zenith_cli: ZenithCli,
+                                   test_output_dir: str,
+                                   pg: Postgres,
+                                   pageserver_pg_port: int):
 
     # Get the timeline ID of our branch. We need it for the 'basebackup' command
     with closing(pg.connect()) as conn:
@@ -1116,7 +1193,7 @@ def check_restored_datadir_content(zenith_cli: ZenithCli, test_output_dir: str,
     pg.stop()
 
     # Take a basebackup from pageserver
-    restored_dir_path = os.path.join(test_output_dir, f"{pg.branch}_restored_datadir")
+    restored_dir_path = os.path.join(test_output_dir, f"{pg.node_name}_restored_datadir")
     mkdir_if_needed(restored_dir_path)
 
     psql_path = os.path.join(pg.pg_bin.pg_bin_path, 'psql')
@@ -1124,7 +1201,7 @@ 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}
     """

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

test_runner/performance/test_bulk_insert.py

@@ -5,6 +5,7 @@ from fixtures.log_helper import log
 
 pytest_plugins = ("fixtures.zenith_fixtures", "fixtures.benchmark_fixture")
 
+
 #
 # Run bulk INSERT test.
 #
@@ -15,7 +16,12 @@ pytest_plugins = ("fixtures.zenith_fixtures", "fixtures.benchmark_fixture")
 # 3. Disk space used
 # 4. Peak memory usage
 #
-def test_bulk_insert(postgres: PostgresFactory, pageserver: ZenithPageserver, pg_bin, zenith_cli, zenbenchmark, repo_dir: str):
+def test_bulk_insert(postgres: PostgresFactory,
+                     pageserver: ZenithPageserver,
+                     pg_bin,
+                     zenith_cli,
+                     zenbenchmark,
+                     repo_dir: str):
     # Create a branch for us
     zenith_cli.run(["branch", "test_bulk_insert", "empty"])
 
@@ -24,7 +30,7 @@ def test_bulk_insert(postgres: PostgresFactory, pageserver: ZenithPageserver, pg
 
     # Open a connection directly to the page server that we'll use to force
     # flushing the layers to disk
-    psconn = pageserver.connect();
+    psconn = pageserver.connect()
     pscur = psconn.cursor()
 
     # Get the timeline ID of our branch. We need it for the 'do_gc' command
@@ -48,5 +54,7 @@ def test_bulk_insert(postgres: PostgresFactory, pageserver: ZenithPageserver, pg
             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')
+            timeline_size = zenbenchmark.get_timeline_size(repo_dir,
+                                                           pageserver.initial_tenant,
+                                                           timeline)
+            zenbenchmark.record('size', timeline_size / (1024 * 1024), 'MB')

test_runner/performance/test_bulk_tenant_create.py

@@ -37,7 +37,9 @@ def test_bulk_tenant_create(
 
         tenant = tenant_factory.create()
         zenith_cli.run([
-            "branch", f"test_bulk_tenant_create_{tenants_count}_{i}_{use_wal_acceptors}", "main",
+            "branch",
+            f"test_bulk_tenant_create_{tenants_count}_{i}_{use_wal_acceptors}",
+            "main",
             f"--tenantid={tenant}"
         ])
 
@@ -46,6 +48,7 @@ def test_bulk_tenant_create(
 
         pg_tenant = postgres.create_start(
             f"test_bulk_tenant_create_{tenants_count}_{i}_{use_wal_acceptors}",
+            None,  # branch name, None means same as node name
             tenant,
             wal_acceptors=wa_factory.get_connstrs() if use_wal_acceptors == 'with_wa' else None,
         )

test_runner/performance/test_gist_build.py

@@ -5,12 +5,18 @@ 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):
+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"])
 
@@ -19,7 +25,7 @@ def test_gist_buffering_build(postgres: PostgresFactory, pageserver: ZenithPages
 
     # Open a connection directly to the page server that we'll use to force
     # flushing the layers to disk
-    psconn = pageserver.connect();
+    psconn = pageserver.connect()
     pscur = psconn.cursor()
 
     # Get the timeline ID of our branch. We need it for the 'do_gc' command
@@ -29,13 +35,17 @@ def test_gist_buffering_build(postgres: PostgresFactory, pageserver: ZenithPages
             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;");
+            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)");
+                    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
@@ -45,5 +55,7 @@ def test_gist_buffering_build(postgres: PostgresFactory, pageserver: ZenithPages
             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')
+            timeline_size = zenbenchmark.get_timeline_size(repo_dir,
+                                                           pageserver.initial_tenant,
+                                                           timeline)
+            zenbenchmark.record('size', timeline_size / (1024 * 1024), 'MB')

test_runner/performance/test_perf_pgbench.py

@@ -5,6 +5,7 @@ from fixtures.log_helper import log
 
 pytest_plugins = ("fixtures.zenith_fixtures", "fixtures.benchmark_fixture")
 
+
 #
 # Run a very short pgbench test.
 #
@@ -14,7 +15,12 @@ pytest_plugins = ("fixtures.zenith_fixtures", "fixtures.benchmark_fixture")
 # 2. Time to run 5000 pgbench transactions
 # 3. Disk space used
 #
-def test_pgbench(postgres: PostgresFactory, pageserver: ZenithPageserver, pg_bin, zenith_cli, zenbenchmark, repo_dir: str):
+def test_pgbench(postgres: PostgresFactory,
+                 pageserver: ZenithPageserver,
+                 pg_bin,
+                 zenith_cli,
+                 zenbenchmark,
+                 repo_dir: str):
     # Create a branch for us
     zenith_cli.run(["branch", "test_pgbench_perf", "empty"])
 
@@ -23,7 +29,7 @@ def test_pgbench(postgres: PostgresFactory, pageserver: ZenithPageserver, pg_bin
 
     # Open a connection directly to the page server that we'll use to force
     # flushing the layers to disk
-    psconn = pageserver.connect();
+    psconn = pageserver.connect()
     pscur = psconn.cursor()
 
     # Get the timeline ID of our branch. We need it for the 'do_gc' command
@@ -53,4 +59,4 @@ def test_pgbench(postgres: PostgresFactory, pageserver: ZenithPageserver, pg_bin
 
     # 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')
+    zenbenchmark.record('size', timeline_size / (1024 * 1024), 'MB')

test_runner/performance/test_write_amplification.py

@@ -17,7 +17,13 @@ from fixtures.log_helper import log
 
 pytest_plugins = ("fixtures.zenith_fixtures", "fixtures.benchmark_fixture")
 
-def test_write_amplification(postgres: PostgresFactory, pageserver: ZenithPageserver, pg_bin, zenith_cli, zenbenchmark, repo_dir: str):
+
+def test_write_amplification(postgres: PostgresFactory,
+                             pageserver: ZenithPageserver,
+                             pg_bin,
+                             zenith_cli,
+                             zenbenchmark,
+                             repo_dir: str):
     # Create a branch for us
     zenith_cli.run(["branch", "test_write_amplification", "empty"])
 
@@ -26,7 +32,7 @@ def test_write_amplification(postgres: PostgresFactory, pageserver: ZenithPagese
 
     # Open a connection directly to the page server that we'll use to force
     # flushing the layers to disk
-    psconn = pageserver.connect();
+    psconn = pageserver.connect()
     pscur = psconn.cursor()
 
     with closing(pg.connect()) as conn:
@@ -71,5 +77,7 @@ def test_write_amplification(postgres: PostgresFactory, pageserver: ZenithPagese
                         pscur.execute(f"do_gc {pageserver.initial_tenant} {timeline} 0")
 
             # 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')
+            timeline_size = zenbenchmark.get_timeline_size(repo_dir,
+                                                           pageserver.initial_tenant,
+                                                           timeline)
+            zenbenchmark.record('size', timeline_size / (1024 * 1024), 'MB')

test_runner/setup.cfg

@@ -10,6 +10,7 @@ max-line-length = 100
 [yapf]
 based_on_style = pep8
 column_limit = 100
+split_all_top_level_comma_separated_values = true
 
 [mypy]
 # some tests don't typecheck when this flag is set

zenith/src/main.rs

@@ -32,12 +32,16 @@ struct BranchTreeEl {
 //   * Providing CLI api to the pageserver
 //   * TODO: export/import to/from usual postgres
 fn main() -> Result<()> {
-    let timeline_arg = Arg::with_name("timeline")
-        .short("n")
+    let node_arg = Arg::with_name("node")
         .index(1)
-        .help("Timeline name")
+        .help("Node name")
         .required(true);
 
+    let timeline_arg = Arg::with_name("timeline")
+        .index(2)
+        .help("Branch name or a point-in time specification")
+        .required(false);
+
     let tenantid_arg = Arg::with_name("tenantid")
         .long("tenantid")
         .help("Tenant id. Represented as a hexadecimal string 32 symbols length")
@@ -102,7 +106,10 @@ fn main() -> Result<()> {
                 .subcommand(SubCommand::with_name("list").arg(tenantid_arg.clone()))
                 .subcommand(SubCommand::with_name("create")
                     .about("Create a postgres compute node")
-                    .arg(timeline_arg.clone()).arg(tenantid_arg.clone()).arg(port_arg.clone())
+                    .arg(node_arg.clone())
+                    .arg(timeline_arg.clone())
+                    .arg(tenantid_arg.clone())
+                    .arg(port_arg.clone())
                     .arg(
                         Arg::with_name("config-only")
                             .help("Don't do basebackup, create compute node with only config files")
@@ -111,13 +118,13 @@ fn main() -> Result<()> {
                     ))
                 .subcommand(SubCommand::with_name("start")
                     .about("Start a postgres compute node.\n This command actually creates new node from scratch, but preserves existing config files")
-                    .arg(
-                        timeline_arg.clone()
-                    ).arg(
-                        tenantid_arg.clone()
-                    ).arg(port_arg.clone()))
+                    .arg(node_arg.clone())
+                    .arg(timeline_arg.clone())
+                    .arg(tenantid_arg.clone())
+                    .arg(port_arg.clone()))
                 .subcommand(
                     SubCommand::with_name("stop")
+                        .arg(node_arg.clone())
                         .arg(timeline_arg.clone())
                         .arg(tenantid_arg.clone())
                         .arg(
@@ -430,25 +437,32 @@ fn handle_pg(pg_match: &ArgMatches, env: &local_env::LocalEnv) -> Result<()> {
             let tenantid: ZTenantId = list_match
                 .value_of("tenantid")
                 .map_or(Ok(env.tenantid), |value| value.parse())?;
+
             let branch_infos = get_branch_infos(env, &tenantid).unwrap_or_else(|e| {
                 eprintln!("Failed to load branch info: {}", e);
                 HashMap::new()
             });
 
-            println!("BRANCH\tADDRESS\t\tLSN\t\tSTATUS");
-            for ((_, timeline_name), node) in cplane
+            println!("NODE\tADDRESS\t\tBRANCH\tLSN\t\tSTATUS");
+            for ((_, node_name), node) in cplane
                 .nodes
                 .iter()
                 .filter(|((node_tenantid, _), _)| node_tenantid == &tenantid)
             {
+                // FIXME: This shows the LSN at the end of the timeline. It's not the
+                // right thing to do for read-only nodes that might be anchored at an
+                // older point in time, or following but lagging behind the primary.
+                let lsn_str = branch_infos
+                    .get(&node.timelineid)
+                    .map(|bi| bi.latest_valid_lsn.to_string())
+                    .unwrap_or_else(|| "?".to_string());
+
                 println!(
-                    "{}\t{}\t{}\t{}",
-                    timeline_name,
+                    "{}\t{}\t{}\t{}\t{}",
+                    node_name,
                     node.address,
-                    branch_infos
-                        .get(&node.timelineid)
-                        .map(|bi| bi.latest_valid_lsn.to_string())
-                        .unwrap_or_else(|| "?".to_string()),
+                    node.timelineid, // FIXME: resolve human-friendly branch name
+                    lsn_str,
                     node.status(),
                 );
             }
@@ -457,26 +471,28 @@ fn handle_pg(pg_match: &ArgMatches, env: &local_env::LocalEnv) -> Result<()> {
             let tenantid: ZTenantId = create_match
                 .value_of("tenantid")
                 .map_or(Ok(env.tenantid), |value| value.parse())?;
-            let timeline_name = create_match.value_of("timeline").unwrap_or("main");
+            let node_name = create_match.value_of("node").unwrap_or("main");
+            let timeline_name = create_match.value_of("timeline").unwrap_or(node_name);
 
             let port: Option<u16> = match create_match.value_of("port") {
                 Some(p) => Some(p.parse()?),
                 None => None,
             };
-            cplane.new_node(tenantid, timeline_name, port)?;
+            cplane.new_node(tenantid, node_name, timeline_name, port)?;
         }
         ("start", Some(start_match)) => {
             let tenantid: ZTenantId = start_match
                 .value_of("tenantid")
                 .map_or(Ok(env.tenantid), |value| value.parse())?;
-            let timeline_name = start_match.value_of("timeline").unwrap_or("main");
+            let node_name = start_match.value_of("node").unwrap_or("main");
+            let timeline_name = start_match.value_of("timeline");
 
             let port: Option<u16> = match start_match.value_of("port") {
                 Some(p) => Some(p.parse()?),
                 None => None,
             };
 
-            let node = cplane.nodes.get(&(tenantid, timeline_name.to_owned()));
+            let node = cplane.nodes.get(&(tenantid, node_name.to_owned()));
 
             let auth_token = if matches!(env.auth_type, AuthType::ZenithJWT) {
                 let claims = Claims::new(Some(tenantid), Scope::Tenant);
@@ -485,12 +501,11 @@ fn handle_pg(pg_match: &ArgMatches, env: &local_env::LocalEnv) -> Result<()> {
                 None
             };
 
-            println!(
-                "Starting {} postgres on timeline {}...",
-                if node.is_some() { "existing" } else { "new" },
-                timeline_name
-            );
             if let Some(node) = node {
+                if timeline_name.is_some() {
+                    println!("timeline name ignored because node exists already");
+                }
+                println!("Starting existing postgres {}...", node_name);
                 node.start(&auth_token)?;
             } else {
                 // when used with custom port this results in non obvious behaviour
@@ -498,12 +513,17 @@ fn handle_pg(pg_match: &ArgMatches, env: &local_env::LocalEnv) -> Result<()> {
                 // start --port X
                 // stop
                 // start <-- will also use port X even without explicit port argument
-                let node = cplane.new_node(tenantid, timeline_name, port)?;
+                let timeline_name = timeline_name.unwrap_or(node_name);
+                println!(
+                    "Starting new postgres {} on {}...",
+                    node_name, timeline_name
+                );
+                let node = cplane.new_node(tenantid, node_name, timeline_name, port)?;
                 node.start(&auth_token)?;
             }
         }
         ("stop", Some(stop_match)) => {
-            let timeline_name = stop_match.value_of("timeline").unwrap_or("main");
+            let node_name = stop_match.value_of("node").unwrap_or("main");
             let destroy = stop_match.is_present("destroy");
             let tenantid: ZTenantId = stop_match
                 .value_of("tenantid")
@@ -511,8 +531,8 @@ fn handle_pg(pg_match: &ArgMatches, env: &local_env::LocalEnv) -> Result<()> {
 
             let node = cplane
                 .nodes
-                .get(&(tenantid, timeline_name.to_owned()))
-                .ok_or_else(|| anyhow!("postgres {} is not found", timeline_name))?;
+                .get(&(tenantid, node_name.to_owned()))
+                .ok_or_else(|| anyhow!("postgres {} is not found", node_name))?;
             node.stop(destroy)?;
         }
 

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)
         );
     }
 }

zenith_utils/src/vec_map.rs

@@ -87,6 +87,10 @@ impl<K: Ord, V> VecMap<K, V> {
         Ok(None)
     }
 
+    pub fn last(&self) -> Option<&(K, V)> {
+        self.0.last()
+    }
+
     /// Split the map into two.
     ///
     /// The left map contains everything before `cutoff` (exclusive).