tests: make Workload more determinstic

test_runner/fixtures/neon_cli.py

@@ -486,6 +486,7 @@ def endpoint_create(
         lsn: Lsn | None = None,
         pageserver_id: int | None = None,
         allow_multiple=False,
+        update_catalog: bool = False,
     ) -> subprocess.CompletedProcess[str]:
         args = [
             "endpoint",
@@ -511,6 +512,8 @@ def endpoint_create(
             args.extend(["--pageserver-id", str(pageserver_id)])
         if allow_multiple:
             args.extend(["--allow-multiple"])
+        if update_catalog:
+            args.extend(["--update-catalog"])
 
         res = self.raw_cli(args)
         res.check_returncode()

test_runner/fixtures/neon_fixtures.py

@@ -3846,6 +3846,7 @@ def create(
         config_lines: list[str] | None = None,
         pageserver_id: int | None = None,
         allow_multiple: bool = False,
+        update_catalog: bool = False,
     ) -> Self:
         """
         Create a new Postgres endpoint.
@@ -3870,6 +3871,7 @@ def create(
             pg_version=self.env.pg_version,
             pageserver_id=pageserver_id,
             allow_multiple=allow_multiple,
+            update_catalog=update_catalog,
         )
         path = Path("endpoints") / self.endpoint_id / "pgdata"
         self.pgdata_dir = self.env.repo_dir / path
@@ -4283,6 +4285,7 @@ def create(
         hot_standby: bool = False,
         config_lines: list[str] | None = None,
         pageserver_id: int | None = None,
+        update_catalog: bool = False,
     ) -> Endpoint:
         ep = Endpoint(
             self.env,
@@ -4303,6 +4306,7 @@ def create(
             hot_standby=hot_standby,
             config_lines=config_lines,
             pageserver_id=pageserver_id,
+            update_catalog=update_catalog,
         )
 
     def stop_all(self, fail_on_error=True) -> Self:

test_runner/fixtures/workload.py

@@ -53,6 +53,8 @@ def __init__(
         self._endpoint: Endpoint | None = None
         self._endpoint_opts = endpoint_opts or {}
 
+        self._configured_pageserver: int | None = None
+
     def branch(
         self,
         timeline_id: TimelineId,
@@ -92,8 +94,12 @@ def endpoint(self, pageserver_id: int | None = None) -> Endpoint:
                     **self._endpoint_opts,
                 )
                 self._endpoint.start(pageserver_id=pageserver_id)
+                self._configured_pageserver = pageserver_id
             else:
-                self._endpoint.reconfigure(pageserver_id=pageserver_id)
+                if self._configured_pageserver != pageserver_id:
+                    self._configured_pageserver = pageserver_id
+                    self._endpoint.reconfigure(pageserver_id=pageserver_id)
+                    self._endpoint_config = pageserver_id
 
         connstring = self._endpoint.safe_psql(
             "SELECT setting FROM pg_settings WHERE name='neon.pageserver_connstring'"
@@ -122,6 +128,7 @@ def init(self, pageserver_id: int | None = None, allow_recreate=False):
 
     def write_rows(self, n: int, pageserver_id: int | None = None, upload: bool = True):
         endpoint = self.endpoint(pageserver_id)
+
         start = self.expect_rows
         end = start + n - 1
         self.expect_rows += n

test_runner/regress/test_compaction.py

@@ -689,9 +689,7 @@ def test_pageserver_compaction_circuit_breaker(neon_env_builder: NeonEnvBuilder)
     env.pageserver.http_client().configure_failpoints((FAILPOINT, "return"))
 
     # Write some data to trigger compaction
-    workload.write_rows(1024, upload=False)
-    workload.write_rows(1024, upload=False)
-    workload.write_rows(1024, upload=False)
+    workload.write_rows(32768, upload=False)
 
     def assert_broken():
         env.pageserver.assert_log_contains(BROKEN_LOG)

test_runner/regress/test_sharding.py

@@ -91,7 +91,7 @@ def get_sizes():
     workload.init()
 
     sizes_before = get_sizes()
-    workload.write_rows(256)
+    workload.write_rows(65536)
 
     # Test that we can read data back from a sharded tenant
     workload.validate()
@@ -1368,6 +1368,7 @@ def test_sharding_split_failures(
     workload = Workload(env, tenant_id, timeline_id)
     workload.init()
     workload.write_rows(100)
+    compute_reconfigure_listener.register_workload(workload)
 
     # Put the environment into a failing state (exact meaning depends on `failure`)
     failure.apply(env)
@@ -1546,6 +1547,9 @@ def shards_info():
                 # Tip: set to 100MB to make the test fail
                 "max_replication_write_lag=1MB",
             ],
+            # We need `neon` extension for calling backpressure functions,
+            # this flag instructs `compute_ctl` to pre-install it.
+            "update_catalog": True,
         },
     )
     workload.init()
@@ -1815,6 +1819,9 @@ def test_sharding_gc(
         # This is not okay, but it's not a scrubber bug: it's a pageserver issue that is exposed by
         # the specific pattern of aggressive checkpointing+image layer generation + GC that this test does.
         # TODO: remove when /~https://github.com/neondatabase/neon/issues/10720 is fixed
-        ps.allowed_errors.append(
-            ".*could not find data for key 020000000000000000000000000000000000.*"
+        ps.allowed_errors.extend(
+            [
+                ".*could not find data for key 020000000000000000000000000000000000.*",
+                ".*could not ingest record.*",
+            ]
         )

test_runner/regress/test_storage_scrubber.py

@@ -316,8 +316,11 @@ def test_scrubber_physical_gc_ancestors(neon_env_builder: NeonEnvBuilder, shard_
         # This is not okay, but it's not a scrubber bug: it's a pageserver issue that is exposed by
         # the specific pattern of aggressive checkpointing+image layer generation + GC that this test does.
         # TODO: remove when /~https://github.com/neondatabase/neon/issues/10720 is fixed
-        ps.allowed_errors.append(
-            ".*could not find data for key 020000000000000000000000000000000000.*"
+        ps.allowed_errors.extend(
+            [
+                ".*could not find data for key 020000000000000000000000000000000000.*",
+                ".*could not ingest record.*",
+            ]
         )