Antalya 26.3: DataLakeCatalog: avoid full catalog read for UNKNOWN_TABLE typo hints

[zvonand]

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).

Cherry-picked from ClickHouse#100452.

---

Documentation entry for user-facing changes

When show_data_lake_catalogs_in_system_tables = 0, the server must not implicitly scan the whole remote data lake catalog.
Previously, building the “Maybe you meant …” hint for a missing table in a DataLakeCatalog database still called getAllTableNames() → DatabaseDataLake::getTablesIterator(), which lists the entire catalog and loads per-table metadata—heavy work and can OOM on large catalogs, only to enrich an error message.

This change makes TableNameHints::getAllRegisteredNames() return an empty name list for data lake catalogs when that setting is off, so hint generation does not trigger a full catalog listing.

Query examples

SET show_data_lake_catalogs_in_system_tables = 0;

Drop non-existent table:

DROP TABLE datalake.`schema1.table1`;

Previously (undesired): server could answer with UNKNOWN_TABLE and a “Maybe you meant …” suggestion after scanning catalog names:

Received exception from server (version 26.3.1):
Code: 60. DB::Exception: Received from localhost:9000. DB::Exception: Table datalake.`schema1.table1` does not exist. Maybe you meant datalake.`schema1.table`?. (UNKNOWN_TABLE)

After the fix: UNKNOWN_TABLE without hint when the setting is 0:

Received exception from server (version 26.4.1):
Code: 60. DB::Exception: Received from localhost:9000. DB::Exception: Table datalake.`schema1.table1` does not exist. (UNKNOWN_TABLE)

Optional follow-up

If local hints are skipped for data lake + setting 0, getHintForTable can still fall back to getExtendedHintForTable, which only scans non–data-lake databases. In edge cases a suggestion could point at another database’s table. If that is undesirable, a follow-up could skip extended hints under the same condition as local enumeration.

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 [9d04f36]

[alsugiliazova]

Verification report: Altinity/ClickHouse PR #1675

Conclusion

PR is merged. CI red on head, but every failure is either a pre-existing flake or a regression-suite scenario already broken at baseline on antalya-26.3. No PR-caused regression found. The change is a 4-line guard against full catalog reads for typo hints — very small surface, very small blast radius.

Caveat — partial frontport (same as #1640 / #1646 / #1687 / #1694 / #1723). Companion antalya-26.1 frontports are still missing on antalya-26.3; chronic regression failures here are branch-level missing-dependency symptoms. Iceberg sort key timezone tests continue to fail with the same UNRECOGNIZED_ARGUMENTS: '--iceberg_partition_timezone' from a binary that doesn't expose the option.

---

CI on head 9d04f36e — failures

PR test workflow (2 failed checks, 42 success)

Check	Test FAIL	Class
Stateless tests (arm_asan, azure, parallel, 2/4)	00084_external_aggregation	Pre-existing flake — 28 / 12 PRs / 90d
Stateless tests (arm_binary, sequential)	00157_cache_dictionary	Pre-existing flake — 106 / 25 PRs / 90d

Regression workflow (10 failed checks)

Check	Top failing tests on PR-1675 builds (30d)	Baseline (antalya-26.3, 30d)	Class
Swarms (Release + Aarch64)	swarm sanity / cluster with only one observer node, swarm sanity / check scale up and down, swarm joins / join clause, swarm union (×4 each)	30–49% on every PR	Pre-existing broken
S3Export (partition) (Release + Aarch64)	sanity / mismatched columns (×4)	50%	Pre-existing broken
Iceberg (1) (Release + Aarch64)	rest catalog / sort key timezone / day transform positive utc offset, rest catalog / swarm / swarm examples (×4 each)	41% / chronic	Missing-dep + pre-existing
Iceberg (2) (Release + Aarch64)	glue catalog / swarm / swarm examples (×4)	chronic	Pre-existing
Parquet (Release + Aarch64)	postgresql/mysql round-trip compression-type variants (×4 each)	~36%	Pre-existing flaky

Regression DB on /PRs/1675/ builds (30d): 699 Fail / 44,007 OK ≈ 1.6%. Every top failure matches the all-PR baseline fail rate on antalya-26.3.

---

Related to PR diff?

PR is a 4-line guard in DataLakeCatalog — avoid scanning the full catalog when computing UNKNOWN_TABLE typo-suggestion hints (1 file changed). Touches only the error-path of getTable* for data-lake catalogs.

Failing test	Diff overlap	Related?
00084_external_aggregation, 00157_cache_dictionary	none	No
swarms / *, parquet / *, s3_export_partition / *	none (data-lake-catalog typo hints ≠ any of these)	No
iceberg / sort key timezone / *	none — failure is UNRECOGNIZED_ARGUMENTS, missing-dep	No
iceberg / swarm / swarm examples (rest + glue)	none — the change is on the typo-hint code path, not the swarm/iceberg integration	No

No failing test intersects the typo-hint code path or fails above the all-PR baseline.

---

Local checkout

cd /Users/alsugilyazova/workspace/altinity-clickhouse/ClickHouse
gh pr checkout 1675 --repo Altinity/ClickHouse
# HEAD: 9d04f36ed7d7adbae5b146f8bee1389450392e1e

[alsugiliazova]

Audit Report — PR #1675

Confirmed Defects

No confirmed defects in reviewed scope.