Antalya 26.6: Make the export task timeout 24 hours instead of 1 hour

[zvonand]

Changelog category (leave one):

• Improvement

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

Make the export task timeout 24 hours instead of 1 hour (#1843 by @arthurpassos).

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-1843). Cherry-picked from #1843.

• Make the export task timeout 24 hours instead of 1 hour #1843 — Make the export task timeout 24 hours instead of 1 hour

[github-actions]

Workflow [PR], commit [45b53b7]

[zvonand]

@blau-ai

[blau-ai]

CI triage — all failures are pre-existing / infra, none caused by this PR

Verdict: 5 failing checks (+ the PR aggregate gate). 0 PR-caused — all are pre-existing base-branch failures or external-infra flakiness.

This PR is a one-line change: it raises the default of export_merge_tree_partition_task_timeout_seconds from 3600 → 86400 and adds doc text (src/Core/Settings.cpp). That cannot break iceberg, swarms, or unrelated stateless tests, and — importantly — a larger timeout can only ever make tasks live longer (potential hangs), never produce the fast AssertionErrors seen below.

Smoking gun: the unrelated backport #2208 ("Respect use_client_time_zone…", touches nothing in export) shows the exact same failing checks — iceberg_2, s3_export_part, swarms, and the same stateless job. Sibling #2210 does too. These suites are red on antalya-26.6 for every PR right now.

---

Per-check breakdown

Regression {aarch64,release} iceberg_2 — 242 / 2178 scenarios failed.
The PR touches no Iceberg code. Identical failure on the unrelated #2208 and on #2210. → Pre-existing base-branch breakage.

Regression {aarch64,release} s3_export_part — 18–19 / 86 scenarios failed:
error handling/pending mutations, pending patch parts, concurrent alter/{before,during,after} export, concurrent alter/during minio interruption — all quick AssertionErrors (1–9s).
This is the only suite where PR-causation was even theoretically plausible (it exercises export partition tasks). But (a) the same scenarios fail on the unrelated #2208 which doesn't touch the timeout, and (b) the change raises a timeout default — it would manifest as hangs, not sub-second assertion failures. → Pre-existing.

Regression {aarch64,release} swarms — 2–3 / 1521 scenarios failed. Unrelated to export. Same on #2208/#2210. → Pre-existing (likely flaky).

Stateless tests (amd_debug, parallel) — final tally Failed: 3, Broken: 3, all infra/timeout, none touching export or the changed setting:

• 03441_deltalake_clickhouse_virtual_columns, 03441_deltalake_clickhouse_public_datasets → [ BROKEN ] with DELTA_KERNEL_ERROR … HTTP error: error sending request fetching the external clickhouse-public-datasets S3 bucket (_delta_log/_last_checkpoint, 10 retries / 180s). External-network flakiness.
• 03595_extract_url_parameters (600s), 00910_zookeeper_test_alter_compression_codecs_long (300s), 00079_array_join_not_used_joined_column (600s), 04070_url_base_setting (600s BROKEN) → hard timeouts on trivial tests that normally run in <1s, consistent with a stalled/overloaded debug server (aggravated by the delta tests hammering external S3). → Flaky/infra.

PR — aggregate rollup gate; red only because the children above are red. Not an independent failure.

---

Suggested next steps

• No code fix is warranted for this PR — nothing here is attributable to the timeout-default change.
• Safe to re-run Stateless tests (amd_debug, parallel) to clear the external-S3 delta flakiness and the timeouts.
• The iceberg_2 / s3_export_part / swarms regression suites are failing across all open PRs on antalya-26.6; they need to be addressed at the base-branch / clickhouse-regression level (or waived for this port), not in Antalya 26.6: Make the export task timeout 24 hours instead of 1 hour #2209.

The diff itself is low-risk and correct: it only changes a default value and documentation. Once the base-branch regression suites are green (or the failures are confirmed as accepted known-fails), this PR should go green on its own merits.

_{🤖 Generated with Claude Code. Evidence: CI logs from run 31785085273; comparison PRs #2208, #2210.}