Stable 25.8: Bump to 25.8.30

[zvonand]

Changelog category (leave one):

• Not for changelog (changelog entry is not required)

CI/CD Options

Exclude tests:

• Fast test
• Integration Tests
• Stateless tests
• Stateful tests
• Performance tests
• Aarch64 tests
• All with ASAN
• All with TSAN
• All with MSAN
• All with UBSAN
• All with Coverage
• All Regression
• Disable CI Cache

Regression jobs to run:

• Fast suites (mostly <1h)
• Aggregate Functions (2h)
• Alter (1.5h)
• Benchmark (30m)
• ClickHouse Keeper (1h)
• Iceberg (2h)
• LDAP (1h)
• OAuth (5m)
• Parquet (1.5h)
• RBAC (1.5h)
• SSL Server (1h)
• S3 (2h)
• S3 Export (2h)
• Swarms (30m)
• Tiered Storage (2h)

[github-actions]

Workflow [PR], commit [dc287f3]

[zvonand]

@blau-ai

[blau-ai]

CI triage — 3 red checks: 1 real & PR-related (trivial fix), 1 policy (DCO), 1 aggregate

Head SHA dc287f37. Everything else is green (Fast test, all other Stateless splits, Stress, all Regression suites). Verdict:

Check	Class	PR-caused?
Stateless tests (amd_binary, old analyzer, s3 storage, DatabaseReplicated, sequential)	Real test failure	Yes — trivial fix below
DCO	Sign-off policy	Not a code issue — needs a human decision
PR	Aggregate gate	Just reflects the two above

---

1. Stateless / DatabaseReplicated — 04296_alter_rename_column_in_memory_rollback (PR-caused, 1-line fix)

Only 1 test failed in the whole suite (575 passed):

04296_alter_rename_column_in_memory_rollback   FAIL   return code: 47
Code: 47. DB::Exception: Missing columns: 'd' while processing:
'SELECT id, d FROM t_rename_rollback ORDER BY id ASC' ... (UNKNOWN_IDENTIFIER)

Root cause — a missing tag on a newly-added test, not a product bug. This test is new to this PR; it was added by backport commit 17ca607 ("Backport ClickHouse#104822 … Fix race between ALTER RENAME COLUMN and merge causing data loss"). It injects the mt_alter_throw_in_start_mutation failpoint into a plain MergeTree table, runs ALTER … RENAME COLUMN, and asserts that after the throw the in-memory metadata was reverted so SELECT d still works. That invariant is specific to the standalone StorageMergeTree::alter code path. Under --database-replicated the RENAME is driven through the replicated DDL/metadata layer instead, so the storage-level in-memory revert isn't the source of truth and SELECT d legitimately fails → the test is simply not applicable in that mode.

The same backport already knows this: every sibling failpoint/rollback test it adds is tagged no-replicated-database, and the DatabaseReplicated case has its own dedicated test (04311). 04296 is the only one that got no-parallel but not no-replicated-database:

04296_alter_rename_column_in_memory_rollback          # Tags: no-parallel                          <-- missing
04308_alter_rename_column_prepared_mutation_cleanup   # Tags: no-parallel, no-replicated-database
04309_alter_rename_column_post_registration_rollback  # Tags: no-parallel, no-replicated-database
04310_alter_rename_column_settings_rollback           # Tags: no-parallel, no-replicated-database
04311_alter_rename_column_replicated_db_rollback      # Tags: no-parallel, zookeeper, no-replicated-database
04312_alter_rename_column_durable_rollback_throw      # Tags: no-parallel, no-replicated-database

This is deterministic (a failpoint, not a race) and confined to the DatabaseReplicated split — the test passed in every other sequential suite (debug, AsyncInsert, ParallelReplicas, distributed-plan). Not flaky, not infra.

Proposed fix (tests/queries/0_stateless/04296_alter_rename_column_in_memory_rollback.sh), matching its siblings:

 #!/usr/bin/env bash
-# Tags: no-parallel
+# Tags: no-parallel, no-replicated-database
 # Tag no-parallel: uses fail points which affect the whole server.

2. DCO (not PR-code-caused — maintainer decision)

There are 87 commits incorrectly signed off … Sign-off not found

These are the upstream v25.8.30 tag commits pulled in by the merge (e.g. ea13bcb, the backport commits, etc.) — they don't carry Signed-off-by trailers. This is the normal state for a stable-bump PR that merges an upstream tag, and is not something to "fix" in code: rewriting sign-offs across 87 commits would rewrite the merged history. This one is a human call — either override/waive the DCO gate for this bump PR (as is customary here) or configure it to skip merge-bump PRs. I won't rewrite those commits.

3. PR (aggregate)

Green once #1 is fixed and DCO is waived — it carries no independent error.

---

Health check

The bump is in good shape: Fast test (6940 passed), Stress, all Regression suites (iceberg/parquet), and every other Stateless split are green. The only functional red is a single mis-tagged new test — a one-line change. The remaining red (DCO) is a sign-off-policy artifact of merging the upstream tag, not a defect.

Delivery: you asked for the fix but didn't say how you'd like it delivered, so I haven't pushed. Want me to commit the one-line tag fix directly to bump/stable-25.8/25.8.30, or open a separate blau/* PR against it? Say the word and I'll do it.

_{— @blau-ai · evidence: praktika result_pr.json for dc287f37, backport commit 17ca607}