Antalya 26.6: Enabled experimental datalake catalogues by-default

[zvonand]

Changelog category (leave one):

• Improvement

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

Enabled experimental datalake catalogues by default (#1743 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)

Cherry-picked from #1743.

---

Documentation entry for user-facing changes

...

[github-actions]

Workflow [PR], commit [2a98367]

[zvonand]

@blau-ai

[blau-ai]

CI triage for #2040

Verdict: none of the failures are caused by this PR. This is a 9-line change that flips three allow_experimental_database_* settings to true and un-comments the matching SettingsChangesHistory entries — and CI confirms it's clean (no settings-history / fast-test mismatch, which is exactly the check that would catch a bad settings edit). Every red check falls into one of three buckets: infra (release-build runners died), cascade (jobs that couldn't find the release binary the dead builds never uploaded), or pre-existing (regression/stateless failures present on the antalya-26.6 base branch, which is itself red on every recent MasterCI run).

Counts: 2 infra, ~10 cascade from the infra failure, ~11 pre-existing / flaky, 0 PR-caused.

---

1. Infra — root cause (2 checks)

Build (amd_release), Build (arm_release)

The self-hosted runner lost communication with the server. Verify the machine is running and has a healthy network connection…

The runners died mid-build (the Run step ends with a null conclusion after 18m/31m). Not a compile error — the diff builds fine.
Next step: re-run the two Build jobs.

2. Cascade from the dead release builds (~10 checks)

Compatibility check (amd/arm_release), Install packages (amd/arm_release), Docker keeper image, Docker server image, SQLLogic test, SQLStorm test, Stress test (arm_release)

All fail within ~1 min because the release binary was never uploaded. Decisive line (SQLLogic/SQLStorm):

ERROR: S3 object not found: altinity-build-artifacts/PRs/2040/…/build_arm_release/clickhouse

(Stress arm_release shows the same missing-artifact download failing via aws s3 cp.)
Next step: none of these need code changes — they should go green automatically once the Build jobs succeed on re-run.

3. Pre-existing — regression feature gaps (5 release + aarch64 counterparts)

Regression … iceberg_1, iceberg_2, s3_export_part, s3_export_partition, swarms

These are feature-support-validation failures — the regression suite expects Antalya features this 26.6 build doesn't include yet:

Swarm object_storage_cluster_join_mode is not supported by this build; the feature is not present in this version.
Iceberg partition timezone is not supported by this build…
Iceberg partition_key/sorting_key in system.tables is not supported…
Export merge tree part / partition is not supported by this build…

This is a suite-ahead-of-build gap at the branch level, unrelated to flipping the experimental DataLake database flags (which don't touch the iceberg table engine, swarms, or s3-export). Expect the same on a base-branch regression run.
The aarch64 parquet / iceberg_2 / s3part variants that show errored (not failed) died on setup with testflows.uexpect ExpectTimeoutError (~1 min) — infra/flaky bootstrap timeouts, also not PR-related.

4. Pre-existing / flaky stateless (1 check, 6 tests)

Stateless tests (amd_debug, distributed plan, s3 storage, parallel) — Failed: 6

03173_set_transformed_partition_pruning
04033_tpc_ds_q54
04033_tpc_ds_q72
04050_text_index_starts_ends_with
04201_consecutive_keys_nullable_findkey
04266_partition_by_function_of_subcolumn

None touch settings, experimental flags, or datalake catalogs (TPC-DS queries, partition pruning, text index, nullable key lookup). Unrelated to the change on this debug/distributed-plan/s3 config.
Next step: re-run to confirm flaky vs. a known base-branch failure; nothing here for this PR to fix.

---

Bottom line

No fix is needed on this PR for correctness of the change itself. To get a green (or fairly-comparable) run: re-run the two release Build jobs — that clears buckets 1 and 2 (roughly 12 of the red checks). The regression and stateless failures (buckets 3–4) are branch-level / suite-version issues that also fail on antalya-26.6 and are out of scope for a settings-default flip; worth confirming against a recent base-branch run before treating any as a merge blocker.

_{Evidence: runner-death annotations on both Build jobs; S3 object not found in SQLLogic/SQLStorm/Stress logs; feature not present in this version validation lines in the Iceberg/S3Export regression logs; base-branch MasterCI red on all recent runs. I can't build/run ClickHouse in this container, so this is log-based triage — the definitive check is a re-run.}

— @blau-ai