Antalya 26.1 Backport of #100452 - DataLakeCatalog: avoid full catalog read for UNKNOWN_TABLE typo hints

[mkmkme]

Changelog category (leave one):

• Improvement

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

Avoid scanning the whole remote data lake catalog for “Maybe you meant …” table hints when show_data_lake_catalogs_in_system_tables is disabled (ClickHouse#100452 by @alsugiliazova).

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 [50a5ed1]

[mkmkme]

AI audit note: This review comment was generated by AI (claude-4.6-sonnet-medium-thinking).

Audit update for PR #1583 (DataLakeCatalog: avoid full catalog read for UNKNOWN_TABLE typo hints):

Confirmed defects:

No confirmed defects in reviewed scope.

Coverage summary:

• Scope reviewed: DatabaseCatalog.cpp TableNameHints::getAllRegisteredNames() (7 added / 3 deleted lines), call-graph from UNKNOWN_TABLE exception paths through hint generation, getExtendedHintForTable cross-DB path, DatabaseNameHints sibling class, setting declaration and guard logic.
• Categories failed: None.
• Categories passed: guard ordering/null-safety, null-context path, normal-DB path unchanged, extended-hint path unaffected, setting=true path, DatabaseNameHints unaffected, thread safety (no new shared mutable state), C++ lifetime safety.
• Assumptions/limits: Static analysis only; getAllTableNames(nullptr) null-context behavior is pre-existing and not audited here; getExtendedHintForTable hardcoding with_datalake_catalogs = false is pre-existing and out of scope.

---

One pre-existing design note (not a defect in this PR): getExtendedHintForTable unconditionally filters out datalake catalogs via GetDatabasesOptions{.with_datalake_catalogs = false} regardless of the show_data_lake_catalogs_in_system_tables setting, so cross-database datalake hints remain suppressed even when a user sets the flag to true. This asymmetry predates PR 1583 and would be a separate improvement if desired.

[alsugiliazova]

PR #1583 CI Verification Report

CI Results Overview

Category	Count
Success	55
Failure	2
Skipped	39
Total (result_pr.json)	96

Dedicated Regression Test: similar_table_names_hint

A dedicated regression test exists for this functionality at iceberg/tests/iceberg_engine/show_data_lake_catalogs_repro.py, scenario similar_table_names_hint. It is included in the iceberg_engine feature (loaded via feature.py line 77) and runs as part of the Iceberg regression suites.

The test:

1. Creates multiple tables with similar names across namespaces in a DataLakeCatalog
2. Verifies that with show_data_lake_catalogs_in_system_tables = 0, catalog tables are hidden from system.tables
3. Verifies that SHOW TABLES FROM still works (explicit listing)
4. Verifies that dropping a non-existent table with a similar name does not show typo hints when show_data_lake_catalogs_in_system_tables = 0 (i.e., the expensive getAllTableNames() is skipped)

Iceberg regression results:

Suite	x86	aarch64
Iceberg 1	Pass (1 module ok, 761 ok scenarios)	Pass
Iceberg 2	Pass	Pass

The similar_table_names_hint test is part of the 761 passing scenarios in the Iceberg 1 suite. Both Iceberg regression suites passed on both platforms.

CI Failures

1. test_graceful_shutdown — Known Issue, Passed on Retry

Job: Integration tests (amd_binary, 2/5)

test_s3_cluster/test.py::test_graceful_shutdown failed (15.4s) but has retry_ok label — passed on retry. Known issue tracked in #1521.

Related to PR: No — Pre-existing known flaky test

2. test_move_after_processing[another_bucket-AzureQueue] — Unrelated Azure Queue Test

Job: Integration tests (arm_binary, distributed plan, 3/4)

1 failure out of 1,377 tests. Azure Queue storage processing test, completely unrelated to DataLakeCatalog hint logic.

Related to PR: No — Azure Queue test

3. Settings Regression (x86 + aarch64) — Snapshot Mismatch

Jobs: RegressionTestsRelease / Common (settings), RegressionTestsAarch64 / Common (settings)

Failed test: /settings/default values/iceberg_metadata_staleness_ms — snapshot baseline mismatch for a newly added setting (from PR #1575). Not related to this PR's change.

Related to PR: No — Snapshot drift from PR #1575's new setting

4. Swarms Regression (x86 + aarch64) — Pre-existing Node Failure Instability

x86: 3 failed + 1 errored scenarios (node failure: restart, network failure, disk space, restart clickhouse)
aarch64: 2 failed + 1 errored scenarios

Pre-existing instability on the antalya-26.1 branch, same pattern as other PRs.

Related to PR: No — Pre-existing Swarms node failure instability

5. AggregateFunctions (3) (aarch64) — Pre-existing Test Failure

Failed test: /aggregate functions/part 3/state/rankCorrState — aggregate function state test. Completely unrelated to DatabaseCatalog.cpp changes.

Related to PR: No — Pre-existing aggregate function issue

6. GrypeScan (-alpine) — CVE in Base Image

CVE-2026-2673 (High) in Alpine base image OpenSSL.

Related to PR: No — Base image vulnerability

Conclusion

Verdict: Ready to merge — No PR-related failures.