[zvonand]

Changelog category (leave one):

• Improvement

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

Fixed ALIAS columns read through Distributed, Merge over Distributed, and Hybrid tables. A JOIN whose two sides declare a same-named ALIAS column failed with MULTIPLE_EXPRESSIONS_FOR_ALIAS; reading an ALIAS column of a Merge table over a Distributed child failed with NOT_FOUND_COLUMN_IN_BLOCK or returned defaults; reading an ALIAS column of a Hybrid table from inside a subquery failed with THERE_IS_NO_COLUMN; and an ALIAS column referenced from inside a lambda over a Distributed table failed on the shard (#2409 by @mkmkme).

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 #2409.

---

This is the attempt of forward-porting #1844 to 26.6 considering the improvements done in the upstream.

Documentation entry for user-facing changes

...

[github-actions]

Workflow [PR], commit [4b8725e]

[zvonand]

@blau-ai

[blau-ai]

CI triage — PR #2662

Verdict: 4 red checks → 0 caused by this PR's source changes. 1 is a stale-branch issue (one merge fixes it), 3 are flaky/infra. SQLLogic has since gone green.

Check	Classification	Cause
Integration tests (arm_binary, distributed plan, 1/4) — 21/2325	Stale branch (not source-related)	Branch is behind base; missing a named-collections privilege backport + its test config
Regression release swarms — 3 scenarios	Flaky / infra	Connection timeouts in resilience tests
Regression aarch64 swarms — 3 scenarios	Flaky / infra	Same scenarios, same cause
SQLLogic test	✅ now passing	(was still running at first look)

---

1. Integration tests (arm_binary, distributed plan, 1/4) — stale branch, fixable by merging base

All 21 failures are in a single file, test_storage_iceberg_with_spark/test_export_partition_iceberg.py, and every one dies at the table-creation setup step (attach_ch_iceberg → create_iceberg_table) with the identical error:

Code: 497. DB::Exception: default: Not enough privileges. To execute this query,
it's necessary to have the grant SHOW NAMED COLLECTIONS SECRETS ON s3.

This is not your alias-marker change:

• The failure happens before any SELECT runs — the tests never reach the distributed/alias-marker code path.
• gh pr diff 2662 touches zero grant / named-collection / iceberg code.
• On a recent base-branch PR (Antalya 26.8: Iceberg: reuse the Puffin object metadata across deletion-vector reads #2665) the same distributed plan, 1/4 group is green.

Root cause is that this branch is ~1 day behind antalya-26.8 and missing the upstream backport ClickHouse#123133 ("Require permission to override named collections"), merged to base via ClickHouse#125982 (commit 64c27aa). That backport makes overriding a named collection require the SHOW NAMED COLLECTIONS SECRETS grant, and in the same change it adds the grant to the iceberg test's users config. On base:

# tests/integration/test_storage_iceberg_with_spark/configs/users.d/users.xml  (present on base, absent here)
             <named_collection_control>1</named_collection_control>
+            <show_named_collections_secrets>1</show_named_collections_secrets>

git merge-base --is-ancestor 64c27aa HEAD → NOT in HEAD, confirming the branch predates it.

Concrete fix: merge origin/antalya-26.8 into the PR branch (git merge origin/antalya-26.8, resolve any conflicts in the analyzer/planner files, push). That brings in both the server-side privilege change and the test-config grant, and the 21 tests should pass. Cherry-picking just the one-line users.xml is not enough — the underlying privilege-check behavior (ClickHouse#123133) must come along too, so take the full merge. (I can do the merge + conflict resolution on request.)

2 & 3. Regression swarms (release + aarch64) — flaky/infra, safe to re-run

Both arches fail the identical 3 leaf scenarios — being arch-independent is itself a strong infra signal:

• cluster discovery / check cluster discovery with wrong cluster name → Code: 279. Cannot connect to any replica for query execution. (ALL_CONNECTION_TRIES_FAILED) on an s3Cluster(...) query.
• node failure / cpu overload → Code: 209 ... Timeout exceeded while reading from socket ... receive timeout 300000 ms (SOCKET_TIMEOUT).
• node failure / cpu overload all swarm nodes → same SOCKET_TIMEOUT family.

These are the swarm resilience suite (deliberate node-kill / CPU-overload fault injection against MinIO/object-storage clusters). The errors are transport-level (connect failures, socket timeouts under induced load) — a planner/alias-marker regression would surface as a logical error or wrong result during analysis, not a 300s socket timeout under CPU overload. 1518/1535 scenarios pass. Classification: flaky/infra — safe to re-run; not PR-related.

---

Bottom line: nothing here points at the __aliasMarker refactoring. Merge antalya-26.8 to clear the integration failures, and re-run the two swarm jobs.

🤖 Generated with Claude Code