Antalya 26.3 Backport of #102113 - Fix LOGICAL_ERROR in Iceberg mutations due to missing datalake_table_state

[mkmkme]

Fix LOGICAL_ERROR in Iceberg mutations due to missing datalake_table_state

Changelog category (leave one):

• Bug Fix (user-visible misbehavior in an official stable release)

Changelog entry (a user-readable short description of the changes that goes to CHANGELOG.md):

Fix server crash (LOGICAL_ERROR) when executing ALTER TABLE UPDATE/DELETE on Iceberg tables when no prior SELECT or INSERT was done on the table in the same server lifetime (ClickHouse#102113 by @alexey-milovidov)

Documentation entry for user-facing changes

...

CI/CD Options

Exclude tests:

• Fast test
• Integration Tests
• Stateless tests
• Stateful tests
• Performance tests
• All with ASAN
• All with TSAN
• All with MSAN
• All with UBSAN
• All with Coverage
• All with Aarch64
• 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)
• Parquet (1.5h)
• RBAC (1.5h)
• SSL Server (1h)
• S3 (2h)
• S3 Export (2h)
• Swarms (30m)
• Tiered Storage (2h)

[github-actions]

Workflow [PR], commit [9f4955e]

[il9ue]

LGTM — clean backport ✅

Diff matches upstream [#102113](ClickHouse#102113) exactly (final merged commit ae2859e), no backport conflicts.

Fix: One-line updateExternalDynamicMetadataIfExists(context_) at the top of StorageObjectStorage::mutate, mirroring the pattern in InterpreterInsertQuery and StorageMaterializedView::read.

Test: Final iteration after several rounds of bugfix-validation upstream — IcebergS3 (not Local, which goes read-only after ATTACH), DETACH/ATTACH to simulate restart, and --validate_mutation_query=0 to avoid MutationsInterpreter::validate masking the bug via its internal InterpreterSelectQuery. Comments capture all of this well.

CI selection looks right: Iceberg, Parquet, S3 Export, Swarms enabled; ASAN kept. Approve once green.

[alsugiliazova]

Verification: PR #1767

Title: Antalya 26.3 Backport of ClickHouse#102113 — Fix LOGICAL_ERROR in Iceberg mutations due to missing datalake_table_state
Base: antalya-26.3 ← backports/antalya-26.3/102113
Head: 9f4955e1c18a2ea943383049377a355c86be2646
Size: +59 / -0 across 3 files · State: OPEN · Mergeable

Changes

• src/Storages/ObjectStorage/StorageObjectStorage.cpp — production fix for missing datalake_table_state during Iceberg mutations (avoids LOGICAL_ERROR when mutating without a prior SELECT).
• tests/queries/0_stateless/04092_iceberg_mutate_without_prior_select.sh — new dedicated stateless test added by this PR.
• tests/queries/0_stateless/04092_iceberg_mutate_without_prior_select.reference — expected output for the new test.

PR-added test

04092_iceberg_mutate_without_prior_select — passes 100% (6/6 runs) across every stateless job that picked it up:

• Stateless tests (amd_asan, distributed plan, parallel, 1/4) — OK
• Stateless tests (amd_debug, distributed plan, s3 storage, parallel) — OK
• Stateless tests (amd_debug, parallel) — OK
• Stateless tests (arm_asan, azure, parallel, 1/4) — OK
• Stateless tests (arm_asan, targeted) — OK
• Stateless tests (arm_binary, parallel) — OK
• Fast test — SKIPPED (expected; not selected by fast-test filter)

The new code path has clean positive coverage.

CI overview (head commit)

• PR test workflow: 44 success / 49 skipped / 7 failure
• Regression workflow: 28 success / 68 skipped / 4 failure (chronic baseline)
• One pending action_required job (queue/auth, not a real failure)

PR test-workflow failures

Check	Verdict
Integration tests (amd_asan, db disk, old analyzer, 3/6) — 24 × test_s3_cluster/test.py::*	Cascade failure, not PR-caused. All 24 tests fail at the same timestamp 2026-05-09 08:30:11 (server-died pattern). The same test_s3_cluster shard cascade is recurring across antalya-26.3 PRs 1773, 1772, 1757, 1741, 1731, 1692 in the last 30 days. The PR diff does not touch s3Cluster, only Iceberg mutation state.
Integration tests (amd_asan, db disk, old analyzer, 2/6)	Job-level error, no test rows in DB.
Stress test (amd_debug) — Test script failed	Job-level (not test) failure; sporadic infra on the branch.

Regression-workflow failures (chronic baseline on antalya-26.3)

Suite	Result
Swarms (Aarch64 + Release)	227 fails — chronic baseline (initiator out of disk space, etc.)
Parquet (Aarch64 + Release)	34 fails — chronic baseline
S3Export partition (Aarch64 + Release)	20 fails — chronic baseline
S3Export part (Aarch64 + Release)	14 fails — chronic baseline

Same fingerprint observed on every recent antalya-26.3 PR. No new failure modes introduced.

Caveat — partial frontport

This PR lands on antalya-26.3 where several companion features from antalya-26.1 are still being frontported in parallel. Some regression suites still rely on settings/CLI args that are not yet wired in this base, so a final re-verify is recommended once the rest of the frontport bundle lands.

Verdict

Safe to merge.

• PR’s own test 04092_iceberg_mutate_without_prior_select passes 100% across all stateless jobs.
• No PR-caused test regressions.
• All failing checks are pre-existing flakes / chronic baseline failures shared with sibling antalya-26.3 PRs.

[alsugiliazova]

Audit: PR #1767 — Antalya 26.3 Backport of #102113 — Fix LOGICAL_ERROR in Iceberg mutations due to missing datalake_table_state

AI audit note: This review comment was generated by AI (Cursor agent, audit-review skill).

Confirmed defects

No confirmed defects in reviewed scope.