Antalya 26.6: Fix file identifier in rescheduleTasksFromReplica

[zvonand]

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):

Solved #1486

File identifier for distributed tasks was changed between 25.8 and 26.1
In frontport #1414 it was missed and rescheduleTasksFromReplica continued to use old variant.

Fix unsyncronized access to replica_to_files_to_be_processed class member (#1493 by @ianton-ru) (#1748 by @zvonand).

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)

Combined port of 1 PR(s) (group pr-1748). Cherry-picked from #1748.

• Antalya 26.3: Fix file identifier in rescheduleTasksFromReplica #1748 — Antalya 26.3: Fix file identifier in rescheduleTasksFromReplica

[github-actions]

Workflow [PR], commit [4bd8d81]

[zvonand]

@blau-ai

[blau-ai]

CI triage — @zvonand

Verdict: none of the failures are caused by this PR. They fall into three pre-existing/infra buckets. The PR's C++ change compiled fine (the stateless/regression jobs ran ~11k tests against build_amd_binary, so there was no build break), and the merge-base build was reused (all Build* jobs show skipping).

Cross-checked against the base branch's own MasterCI on antalya-26.6 (run 31713806632) — it fails on the identical set of Stateless + Regression jobs, so these are branch-level failures, not regressions from this PR.

---

1. Regression: Swarms / S3Export (part, partition) / Iceberg (1, 2) — aarch64 & release — pre-existing (feature-gated)

All fail immediately at feature support validation, before any test logic runs, because the clickhouse-regression suite expects features that the 26.6.2.20000.altinityantalya build doesn't expose:

✘ '/swarms/feature support validation' ᐅ Swarm object_storage_cluster is not supported by this build; the feature is not present in this version.
✘ '/s3/minio/export tests/export part/feature support validation' ᐅ Export merge tree part is not supported by this build; the feature is not present in this version.
✘ '.../iceberg iterator race condition/feature support validation' ᐅ Swarm object_storage_cluster_join_mode is not supported by this build; the feature is not present in this version.
✘ '.../sort key timezone/feature support validation' ᐅ Iceberg partition timezone is not supported by this build; the feature is not present in this version.

(Iceberg 1&2 otherwise pass — 40 ok / 4 failed features, only the 2 feature-gate scenarios fail.)

Not PR-caused. Same jobs fail on the base-branch MasterCI. This is a regression-suite ↔ build version mismatch, not a code defect. A re-run won't fix it; it needs the suite pinned to a matching feature set (regression-team / release-eng concern).

⚠️ Note for you specifically: because those swarm/s3-export/iceberg suites are feature-gated off, the code path this PR actually fixes (object_storage_cluster stable task distribution) isn't being exercised by any green CI signal here. The fix looks correct by inspection, but CI on this branch can't positively validate it.

2. Stateless tests (amd_debug parallel; amd_debug distributed-plan s3-storage parallel; arm_binary parallel) — pre-existing / flaky-infra

Failing tests (same list appears on the base-branch amd_debug job):

Test	Status	Nature
04337_iceberg_v3_row_lineage_reserved_field_id	FAIL (all 3 jobs)	.reference mismatch; upstream test from CH#107377, single-node, unrelated to cluster distribution
03441_deltalake_clickhouse_virtual_columns	BROKEN	pulls external public datasets
03441_deltalake_clickhouse_public_datasets	BROKEN	pulls external public datasets
04070_url_base_setting	BROKEN	url/network
01509_parallel_quorum_insert_no_replicas_long	FAIL (60s)	long/timeout-prone
00046_uniq_upto_distributed, 00071_merge_tree_optimize_aio	FAIL/flaky	—

Not PR-caused. This PR touches only StorageObjectStorageStableTaskDistributor.{cpp,h} (object-storage cluster task distribution). None of these tests exercise that code — they're single-node stateless tests. The same failures (incl. 04337, both 03441, 04070) reproduce on the base branch. Safe to re-run for the genuinely flaky ones; 04337 and the 03441 deltalake tests are deterministic branch-level failures that need a separate fix on antalya-26.6.

3. Grype Scan (server image) — infra (base-image CVEs)

3 high/critical CVEs in the Debian 13 base image OS packages, not in ClickHouse code:

CVE-2026-5450  Critical
CVE-2026-5928  High
CVE-2026-5435  High

Not PR-caused — a source change to a task distributor cannot affect base-image package CVEs. Fix is a base-image bump / vuln allowlist (release-eng), independent of this PR.

4. PR (aggregate gate)

Red only because 1–3 are red; not a standalone failure.

---

Bottom line: 0 PR-caused failures, nothing to fix in this PR. It's a clean forward-port of the rescheduleTasksFromReplica file-identifier fix (+ mutex guarding). The only caveat is coverage, not correctness: the swarm/iceberg regression suites that would exercise the change are feature-gated off on this build, so treat CI here as "no regressions" rather than "fix validated." If you want positive validation, that'd need the object-storage-cluster regression features enabled for the 26.6 antalya build.

_{🤖 evidence-first triage; I can't build/run ClickHouse in this container, so classifications are from CI logs + base-branch comparison.}