Compute release 2025-05-01 11:10 UTC

ARM computes are incoming and we need to account for that in remote
extensions. Previously, we just blindly assumed that all computes were
x86_64.

Note that we use the Go architecture naming convention instead of the
Rust one directly to do our best and be consistent across the stack.

Part-of: https://github.com/neondatabase/cloud/issues/23148

Signed-off-by: Tristan Partin <tristan@neon.tech>

compute/compute-node.Dockerfile

@@ -1096,23 +1096,6 @@ RUN cargo install --locked --version 0.14.1 cargo-pgrx && \
     /bin/bash -c 'cargo pgrx init --pg${PG_VERSION:1}=/usr/local/pgsql/bin/pg_config'
 
 USER root
-#########################################################################################
-#
-# Layer "rust extensions pgrx14"
-#
-# Version 14 is now required by a few 
-# This layer should be used as a base for new pgrx extensions,
-# and eventually get merged with `rust-extensions-build`
-#
-#########################################################################################
-FROM pg-build-nonroot-with-cargo AS rust-extensions-build-pgrx14
-ARG PG_VERSION
-
-RUN cargo install --locked --version 0.14.1 cargo-pgrx && \
-    /bin/bash -c 'cargo pgrx init --pg${PG_VERSION:1}=/usr/local/pgsql/bin/pg_config'
-
-USER root
-
 #########################################################################################
 #
 # Layers "pg-onnx-build" and "pgrag-build"
@@ -1347,39 +1330,6 @@ COPY --from=pg_session_jwt-src /ext-src/ /ext-src/
 WORKDIR /ext-src/pg_session_jwt-src
 RUN cargo pgrx install --release
 
-#########################################################################################
-#
-# Layer "pg-anon-pg-build"
-# compile anon extension
-#
-#########################################################################################
-FROM pg-build AS pg_anon-src
-ARG PG_VERSION
-COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
-WORKDIR /ext-src
-COPY compute/patches/anon_v2.patch .
-
-# This is an experimental extension, never got to real production.
-# !Do not remove! It can be present in shared_preload_libraries and compute will fail to start if library is not found.
-ENV PATH="/usr/local/pgsql/bin/:$PATH"
-RUN wget https://gitlab.com/dalibo/postgresql_anonymizer/-/archive/latest/postgresql_anonymizer-latest.tar.gz -O pg_anon.tar.gz && \
-    mkdir pg_anon-src && cd pg_anon-src && tar xzf ../pg_anon.tar.gz --strip-components=1 -C . && \
-    find /usr/local/pgsql -type f | sed 's|^/usr/local/pgsql/||' > /before.txt && \
-    sed -i 's/pgrx = "0.14.1"/pgrx = { version = "=0.14.1", features = [ "unsafe-postgres" ] }/g' Cargo.toml && \
-    patch -p1 < /ext-src/anon_v2.patch
-
-FROM rust-extensions-build-pgrx14 AS pg-anon-pg-build
-ARG PG_VERSION
-COPY --from=pg_anon-src /ext-src/ /ext-src/
-WORKDIR /ext-src
-RUN cd pg_anon-src && \
-    make -j $(getconf _NPROCESSORS_ONLN) extension PG_CONFIG=/usr/local/pgsql/bin/pg_config PGVER=pg$(echo "$PG_VERSION" | sed 's/^v//') && \
-    make -j $(getconf _NPROCESSORS_ONLN) install PG_CONFIG=/usr/local/pgsql/bin/pg_config PGVER=pg$(echo "$PG_VERSION" | sed 's/^v//') && \
-    chmod -R a+r ../pg_anon-src && \
-    echo 'trusted = true' >> /usr/local/pgsql/share/extension/anon.control;
-
-########################################################################################
-
 #########################################################################################
 #
 # Layer "wal2json-build"
@@ -1676,7 +1626,6 @@ COPY --from=pg_uuidv7-build /usr/local/pgsql/ /usr/local/pgsql/
 COPY --from=pg_roaringbitmap-build /usr/local/pgsql/ /usr/local/pgsql/
 COPY --from=pg_semver-build /usr/local/pgsql/ /usr/local/pgsql/
 COPY --from=wal2json-build /usr/local/pgsql /usr/local/pgsql
-COPY --from=pg-anon-pg-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_mooncake-build /usr/local/pgsql/ /usr/local/pgsql/

compute/patches/anon_v2.patch

@@ -1,129 +0,0 @@
-diff --git a/sql/anon.sql b/sql/anon.sql
-index 0cdc769..f6cc950 100644
---- a/sql/anon.sql
-+++ b/sql/anon.sql
-@@ -1141,3 +1141,8 @@ $$
- -- TODO : https://en.wikipedia.org/wiki/L-diversity
- 
- -- TODO : https://en.wikipedia.org/wiki/T-closeness
-+
-+-- NEON Patches
-+
-+GRANT ALL ON SCHEMA anon to neon_superuser;
-+GRANT ALL ON ALL TABLES IN SCHEMA anon TO neon_superuser;
-diff --git a/sql/init.sql b/sql/init.sql
-index 7da6553..9b6164b 100644
---- a/sql/init.sql
-+++ b/sql/init.sql
-@@ -74,50 +74,49 @@ $$
- 
- SECURITY LABEL FOR anon ON FUNCTION anon.load_csv IS 'UNTRUSTED';
- 
---- load fake data from a given path
--CREATE OR REPLACE FUNCTION anon.init(
--  datapath TEXT
--)
-+CREATE OR REPLACE FUNCTION anon.load_fake_data()
- RETURNS BOOLEAN
- AS $$
- DECLARE
--  datapath_check TEXT;
-   success BOOLEAN;
-+  sharedir TEXT;
-+  datapath TEXT;
- BEGIN
- 
--  IF anon.is_initialized() THEN
--    RAISE NOTICE 'The anon extension is already initialized.';
--    RETURN TRUE;
--  END IF;
-+  datapath := '/extension/anon/';
-+  -- find the local extension directory
-+  SELECT setting INTO sharedir
-+  FROM pg_catalog.pg_config
-+  WHERE name = 'SHAREDIR';
- 
-   SELECT bool_or(results) INTO success
-   FROM unnest(array[
--    anon.load_csv('anon.identifiers_category',datapath||'/identifiers_category.csv'),
--    anon.load_csv('anon.identifier',datapath ||'/identifier.csv'),
--    anon.load_csv('anon.address',datapath ||'/address.csv'),
--    anon.load_csv('anon.city',datapath ||'/city.csv'),
--    anon.load_csv('anon.company',datapath ||'/company.csv'),
--    anon.load_csv('anon.country',datapath ||'/country.csv'),
--    anon.load_csv('anon.email', datapath ||'/email.csv'),
--    anon.load_csv('anon.first_name',datapath ||'/first_name.csv'),
--    anon.load_csv('anon.iban',datapath ||'/iban.csv'),
--    anon.load_csv('anon.last_name',datapath ||'/last_name.csv'),
--    anon.load_csv('anon.postcode',datapath ||'/postcode.csv'),
--    anon.load_csv('anon.siret',datapath ||'/siret.csv'),
--    anon.load_csv('anon.lorem_ipsum',datapath ||'/lorem_ipsum.csv')
-+    anon.load_csv('anon.identifiers_category',sharedir || datapath || '/identifiers_category.csv'),
-+    anon.load_csv('anon.identifier',sharedir || datapath || '/identifier.csv'),
-+    anon.load_csv('anon.address',sharedir || datapath || '/address.csv'),
-+    anon.load_csv('anon.city',sharedir || datapath || '/city.csv'),
-+    anon.load_csv('anon.company',sharedir || datapath || '/company.csv'),
-+    anon.load_csv('anon.country',sharedir || datapath || '/country.csv'),
-+    anon.load_csv('anon.email', sharedir || datapath || '/email.csv'),
-+    anon.load_csv('anon.first_name',sharedir || datapath || '/first_name.csv'),
-+    anon.load_csv('anon.iban',sharedir || datapath || '/iban.csv'),
-+    anon.load_csv('anon.last_name',sharedir || datapath || '/last_name.csv'),
-+    anon.load_csv('anon.postcode',sharedir || datapath || '/postcode.csv'),
-+    anon.load_csv('anon.siret',sharedir || datapath || '/siret.csv'),
-+    anon.load_csv('anon.lorem_ipsum',sharedir || datapath || '/lorem_ipsum.csv')
-   ]) results;
-   RETURN success;
--
- END;
- $$
--  LANGUAGE PLPGSQL
-+  LANGUAGE plpgsql
-   VOLATILE
-   RETURNS NULL ON NULL INPUT
--  PARALLEL UNSAFE -- because load_csv is unsafe
--  SECURITY INVOKER
-+  PARALLEL UNSAFE -- because of the EXCEPTION
-+  SECURITY DEFINER
-   SET search_path=''
- ;
--SECURITY LABEL FOR anon ON FUNCTION anon.init(TEXT) IS 'UNTRUSTED';
-+
-+SECURITY LABEL FOR anon ON FUNCTION anon.load_fake_data IS 'UNTRUSTED';
- 
- -- People tend to forget the anon.init() step
- -- This is a friendly notice for them
-@@ -144,7 +143,7 @@ SECURITY LABEL FOR anon ON FUNCTION anon.notice_if_not_init IS 'UNTRUSTED';
- CREATE OR REPLACE FUNCTION anon.load(TEXT)
- RETURNS BOOLEAN AS
- $$
--  SELECT anon.init($1);
-+  SELECT anon.init();
- $$
-   LANGUAGE SQL
-   VOLATILE
-@@ -159,16 +158,16 @@ SECURITY LABEL FOR anon ON FUNCTION anon.load(TEXT) IS 'UNTRUSTED';
- CREATE OR REPLACE FUNCTION anon.init()
- RETURNS BOOLEAN
- AS $$
--  WITH conf AS (
--        -- find the local extension directory
--        SELECT setting AS sharedir
--        FROM pg_catalog.pg_config
--        WHERE name = 'SHAREDIR'
--    )
--  SELECT anon.init(conf.sharedir || '/extension/anon/')
--  FROM conf;
-+BEGIN
-+  IF anon.is_initialized() THEN
-+    RAISE NOTICE 'The anon extension is already initialized.';
-+    RETURN TRUE;
-+  END IF;
-+
-+  RETURN anon.load_fake_data();
-+END;
- $$
--  LANGUAGE SQL
-+  LANGUAGE plpgsql
-   VOLATILE
-   PARALLEL UNSAFE -- because init is unsafe
-   SECURITY INVOKER

pageserver/ctl/src/layers.rs

@@ -10,7 +10,6 @@ use pageserver::tenant::storage_layer::{DeltaLayer, ImageLayer, delta_layer, ima
 use pageserver::tenant::{TENANTS_SEGMENT_NAME, TIMELINES_SEGMENT_NAME};
 use pageserver::virtual_file::api::IoMode;
 use pageserver::{page_cache, virtual_file};
-use pageserver_api::key::Key;
 use utils::id::{TenantId, TimelineId};
 
 use crate::layer_map_analyzer::parse_filename;
@@ -28,7 +27,6 @@ pub(crate) enum LayerCmd {
         path: PathBuf,
         tenant: String,
         timeline: String,
-        key: Option<Key>,
     },
     /// Dump all information of a layer file
     DumpLayer {
@@ -102,7 +100,6 @@ pub(crate) async fn main(cmd: &LayerCmd) -> Result<()> {
             path,
             tenant,
             timeline,
-            key,
         } => {
             let timeline_path = path
                 .join(TENANTS_SEGMENT_NAME)
@@ -110,37 +107,21 @@ pub(crate) async fn main(cmd: &LayerCmd) -> Result<()> {
                 .join(TIMELINES_SEGMENT_NAME)
                 .join(timeline);
             let mut idx = 0;
-            let mut to_print = Vec::default();
             for layer in fs::read_dir(timeline_path)? {
                 let layer = layer?;
                 if let Ok(layer_file) = parse_filename(&layer.file_name().into_string().unwrap()) {
-                    if let Some(key) = key {
-                        if layer_file.key_range.start <= *key && *key < layer_file.key_range.end {
-                            to_print.push((idx, layer_file));
-                        }
-                    } else {
-                        to_print.push((idx, layer_file));
-                    }
+                    println!(
+                        "[{:3}]  key:{}-{}\n       lsn:{}-{}\n       delta:{}",
+                        idx,
+                        layer_file.key_range.start,
+                        layer_file.key_range.end,
+                        layer_file.lsn_range.start,
+                        layer_file.lsn_range.end,
+                        layer_file.is_delta,
+                    );
                     idx += 1;
                 }
             }
-
-            if key.is_some() {
-                to_print
-                    .sort_by_key(|(_idx, layer_file)| std::cmp::Reverse(layer_file.lsn_range.end));
-            }
-
-            for (idx, layer_file) in to_print {
-                println!(
-                    "[{:3}]  key:{}-{}\n       lsn:{}-{}\n       delta:{}",
-                    idx,
-                    layer_file.key_range.start,
-                    layer_file.key_range.end,
-                    layer_file.lsn_range.start,
-                    layer_file.lsn_range.end,
-                    layer_file.is_delta,
-                );
-            }
             Ok(())
         }
         LayerCmd::DumpLayer {

storage_controller/src/service.rs

@@ -5181,8 +5181,7 @@ impl Service {
             }
 
             // We don't expect any new_shard_count shards to exist here, but drop them just in case
-            tenants
-                .retain(|id, s| !(id.tenant_id == *tenant_id && s.shard.count == *new_shard_count));
+            tenants.retain(|_id, s| s.shard.count != *new_shard_count);
 
             detach_locations
         };

test_runner/fixtures/neon_fixtures.py

@@ -1279,8 +1279,7 @@ class NeonEnv:
                 )
 
             tenant_config = ps_cfg.setdefault("tenant_config", {})
-            # This feature is pending rollout.
-            # tenant_config["rel_size_v2_enabled"] = True
+            tenant_config["rel_size_v2_enabled"] = True  # Enable relsize_v2 by default in tests
 
             if self.pageserver_remote_storage is not None:
                 ps_cfg["remote_storage"] = remote_storage_to_toml_dict(

test_runner/regress/test_attach_tenant_config.py

@@ -186,7 +186,7 @@ def test_fully_custom_config(positive_env: NeonEnv):
             "type": "interpreted",
             "args": {"format": "bincode", "compression": {"zstd": {"level": 1}}},
         },
-        "rel_size_v2_enabled": True,
+        "rel_size_v2_enabled": False,  # test suite enables it by default as of https://github.com/neondatabase/neon/issues/11081, so, custom config means disabling it
         "gc_compaction_enabled": True,
         "gc_compaction_verification": False,
         "gc_compaction_initial_threshold_kb": 1024000,

test_runner/regress/test_compaction.py

@@ -229,7 +229,7 @@ def test_pageserver_gc_compaction_preempt(
 
 
 @skip_in_debug_build("only run with release build")
-@pytest.mark.timeout(900)  # This test is slow with sanitizers enabled, especially on ARM
+@pytest.mark.timeout(600)  # This test is slow with sanitizers enabled, especially on ARM
 @pytest.mark.parametrize(
     "with_branches",
     ["with_branches", "no_branches"],

test_runner/regress/test_pg_regress.py

@@ -471,7 +471,7 @@ def test_tx_abort_with_many_relations(
         try:
             # Rollback phase should be fast: this is one WAL record that we should process efficiently
             fut = exec.submit(rollback_and_wait)
-            fut.result(timeout=15 if reldir_type == "v1" else 30)
+            fut.result(timeout=15)
         except:
             exec.shutdown(wait=False, cancel_futures=True)
             raise

test_runner/regress/test_sharding.py

@@ -1334,13 +1334,6 @@ def test_sharding_split_failures(
         tenant_id, timeline_id, shard_count=initial_shard_count, placement_policy='{"Attached":1}'
     )
 
-    # Create bystander tenants with various shard counts. They should not be affected by the aborted
-    # splits. Regression test for https://github.com/neondatabase/cloud/issues/28589.
-    bystanders = {}  # id → shard_count
-    for bystander_shard_count in [1, 2, 4, 8]:
-        id, _ = env.create_tenant(shard_count=bystander_shard_count)
-        bystanders[id] = bystander_shard_count
-
     env.storage_controller.allowed_errors.extend(
         [
             # All split failures log a warning when then enqueue the abort operation
@@ -1401,8 +1394,6 @@ def test_sharding_split_failures(
             locations = ps.http_client().tenant_list_locations()["tenant_shards"]
             for loc in locations:
                 tenant_shard_id = TenantShardId.parse(loc[0])
-                if tenant_shard_id.tenant_id != tenant_id:
-                    continue  # skip bystanders
                 log.info(f"Shard {tenant_shard_id} seen on node {ps.id} in mode {loc[1]['mode']}")
                 assert tenant_shard_id.shard_count == initial_shard_count
                 if loc[1]["mode"] == "Secondary":
@@ -1423,8 +1414,6 @@ def test_sharding_split_failures(
             locations = ps.http_client().tenant_list_locations()["tenant_shards"]
             for loc in locations:
                 tenant_shard_id = TenantShardId.parse(loc[0])
-                if tenant_shard_id.tenant_id != tenant_id:
-                    continue  # skip bystanders
                 log.info(f"Shard {tenant_shard_id} seen on node {ps.id} in mode {loc[1]['mode']}")
                 assert tenant_shard_id.shard_count == split_shard_count
                 if loc[1]["mode"] == "Secondary":
@@ -1507,12 +1496,6 @@ def test_sharding_split_failures(
     # the scheduler reaches an idle state
     env.storage_controller.reconcile_until_idle(timeout_secs=30)
 
-    # Check that all bystanders are still around.
-    for bystander_id, bystander_shard_count in bystanders.items():
-        response = env.storage_controller.tenant_describe(bystander_id)
-        assert TenantId(response["tenant_id"]) == bystander_id
-        assert len(response["shards"]) == bystander_shard_count
-
     env.storage_controller.consistency_check()