WIP: Add pg_tracing extension

Before we specified the JWT via `SAFEKEEPER_AUTH_TOKEN`, but env vars
are quite public, both in procfs as well as the unit files. So add a way
to put the auth token into a file directly.

context: https://neondb.slack.com/archives/C033RQ5SPDH/p1743692566311099

compute/compute-node.Dockerfile

@@ -1388,6 +1388,38 @@ RUN make -j $(getconf _NPROCESSORS_ONLN) && \
     make -j $(getconf _NPROCESSORS_ONLN) install && \
     echo 'trusted = true' >> /usr/local/pgsql/share/extension/pg_partman.control
 
+#########################################################################################
+# Layer "pg_tracing"
+# compile pg_tracing extension
+#
+#########################################################################################
+FROM build-deps AS pg_tracing-src
+ARG PG_VERSION
+WORKDIR /ext-src
+RUN case "${PG_VERSION:?}" in \
+      "v14" | "v15") \
+        echo "pg_tracing not supported on this PostgreSQL version." && exit 0 \
+	;; \
+      *) \
+      ;; \
+    esac && \
+    wget https://github.com/DataDog/pg_tracing/archive/refs/tags/v0.1.3.tar.gz -O pg_tracing.tar.gz && \
+    echo "d0a7cca7279bb29601ba6c4c1aaeb3a44d71e6afa3b78aae1e3b7269e688f907 pg_tracing.tar.gz" | sha256sum --check && \
+    mkdir pg_tracing-src && cd pg_tracing-src && tar xzf ../pg_tracing.tar.gz --strip-components=1 -C .
+
+FROM pg-build AS pg_tracing-build
+COPY --from=pg_tracing-src /ext-src/ /ext-src/
+WORKDIR /ext-src/pg_tracing-src
+RUN case "${PG_VERSION:?}" in \
+      "v14" | "v15") \
+        echo "pg_tracing not supported on this PostgreSQL version." && exit 0 \
+        ;; \
+      *) \
+      ;; \
+    esac && \
+    make -j $(getconf _NPROCESSORS_ONLN) && \
+    make -j $(getconf _NPROCESSORS_ONLN) install
+
 #########################################################################################
 #
 # Layer "pg_mooncake"
@@ -1617,6 +1649,7 @@ COPY --from=pg_semver-build /usr/local/pgsql/ /usr/local/pgsql/
 COPY --from=wal2json-build /usr/local/pgsql /usr/local/pgsql
 COPY --from=pg_ivm-build /usr/local/pgsql/ /usr/local/pgsql/
 COPY --from=pg_partman-build /usr/local/pgsql/ /usr/local/pgsql/
+COPY --from=pg_tracing-build /usr/local/pgsql/ /usr/local/pgsql/
 COPY --from=pg_mooncake-build /usr/local/pgsql/ /usr/local/pgsql/
 COPY --from=pg_duckdb-build /usr/local/pgsql/ /usr/local/pgsql/
 COPY --from=pg_repack-build /usr/local/pgsql/ /usr/local/pgsql/
@@ -1793,6 +1826,7 @@ COPY --from=pg_semver-src /ext-src/ /ext-src/
 #COPY --from=wal2json-src /ext-src/ /ext-src/
 COPY --from=pg_ivm-src /ext-src/ /ext-src/
 COPY --from=pg_partman-src /ext-src/ /ext-src/
+COPY --from=pg_tracing-src /ext-src/ /ext-src/
 #COPY --from=pg_mooncake-src /ext-src/ /ext-src/
 COPY --from=pg_repack-src /ext-src/ /ext-src/
 COPY --from=pg_repack-build /usr/local/pgsql/ /usr/local/pgsql/

pageserver/src/tenant/timeline/compaction.rs

@@ -1244,6 +1244,10 @@ impl Timeline {
         let mut replace_image_layers = Vec::new();
 
         for layer in layers_to_rewrite {
+            if self.cancel.is_cancelled() {
+                return Err(CompactionError::ShuttingDown);
+            }
+
             tracing::info!(layer=%layer, "Rewriting layer after shard split...");
             let mut image_layer_writer = ImageLayerWriter::new(
                 self.conf,

pgxn/neon/pagestore_smgr.c

@@ -1900,7 +1900,6 @@ neon_wallog_pagev(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
 		log_pages = true;
 	}
 	else if (XLogInsertAllowed() &&
-			 !ShutdownRequestPending &&
 			 (forknum == FSM_FORKNUM || forknum == VISIBILITYMAP_FORKNUM))
 	{
 		log_pages = true;

safekeeper/src/bin/safekeeper.rs

@@ -223,6 +223,9 @@ struct Args {
     /// Flag to use https for requests to peer's safekeeper API.
     #[arg(long)]
     pub use_https_safekeeper_api: bool,
+    /// Path to the JWT auth token used to authenticate with other safekeepers.
+    #[arg(long)]
+    auth_token_path: Option<Utf8PathBuf>,
 }
 
 // Like PathBufValueParser, but allows empty string.
@@ -341,14 +344,24 @@ async fn main() -> anyhow::Result<()> {
     };
 
     // Load JWT auth token to connect to other safekeepers for pull_timeline.
+    // First check if the env var is present, then check the arg with the path.
+    // We want to deprecate and remove the env var method in the future.
     let sk_auth_token = match var("SAFEKEEPER_AUTH_TOKEN") {
         Ok(v) => {
             info!("loaded JWT token for authentication with safekeepers");
             Some(SecretString::from(v))
         }
         Err(VarError::NotPresent) => {
-            info!("no JWT token for authentication with safekeepers detected");
-            None
+            if let Some(auth_token_path) = args.auth_token_path.as_ref() {
+                info!(
+                    "loading JWT token for authentication with safekeepers from {auth_token_path}"
+                );
+                let auth_token = tokio::fs::read_to_string(auth_token_path).await?;
+                Some(SecretString::from(auth_token.trim().to_owned()))
+            } else {
+                info!("no JWT token for authentication with safekeepers detected");
+                None
+            }
         }
         Err(_) => {
             warn!("JWT token for authentication with safekeepers is not unicode");

test_runner/regress/test_wal_acceptor.py

@@ -488,13 +488,14 @@ def test_wal_backup(neon_env_builder: NeonEnvBuilder):
     assert_prefix_empty(neon_env_builder.safekeepers_remote_storage, prefix)
 
 
-def test_s3_wal_replay(neon_env_builder: NeonEnvBuilder):
+# This test is flaky, probably because PUTs of local fs storage are not atomic.
+# Let's keep both remote storage kinds for a while to see if this is the case.
+# https://github.com/neondatabase/neon/issues/10761
+@pytest.mark.parametrize("remote_storage_kind", [s3_storage(), RemoteStorageKind.LOCAL_FS])
+def test_s3_wal_replay(neon_env_builder: NeonEnvBuilder, remote_storage_kind: RemoteStorageKind):
     neon_env_builder.num_safekeepers = 3
 
-    # Note: local fs implementation is not protected from concurrent uploads of
-    # the same segment by different safekeepers and makes the test flaky.
-    # https://github.com/neondatabase/neon/issues/10761
-    neon_env_builder.enable_safekeeper_remote_storage(s3_storage())
+    neon_env_builder.enable_safekeeper_remote_storage(remote_storage_kind)
 
     env = neon_env_builder.init_start()
     tenant_id = env.initial_tenant

vendor/revisions.json

@@ -1,18 +1,18 @@
 {
   "v17": [
     "17.4",
-    "c9e4ff5a38907acd71107634055bf2609aba43a5"
+    "66114c23bc61205b0e3fb1e77ee76a4abc1eb4b8"
   ],
   "v16": [
     "16.8",
-    "746bd9ffe5c29bce030eaea1031054057f3c5d45"
+    "d56e79cd5d6136c159b1d8d98acb7981d4b69364"
   ],
   "v15": [
     "15.12",
-    "23708b3aca9adf163aa0973eb63d9afc0e4a04c3"
+    "aeb292eeace9072e07071254b6ffc7a74007d4d2"
   ],
   "v14": [
     "14.17",
-    "8cca70c22e2894dd4645f9a940086ac437b0a11b"
+    "a0391901a2af13aa029b905272a5b2024133c926"
   ]
 }