25.8.16 Stable Backport of #90059: Unblock ttl part drops for cold volumes

[zvonand]

Changelog category (leave one):

• Bug Fix (user-visible misbehavior in an official stable release)

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

Split part ranges by volume characteristics to enable TTL drop merges for cold volumes. After this patch, parts with a max TTL < now will be removed from cold storage. The algorithm will schedule only single part drops. (ClickHouse#90059 by @Michicosun)

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)
• Tiered Storage (2h)

[github-actions]

Workflow [PR], commit [17cd534]

[zvonand]

@DimensionWieldr I rebased this PR onto 25.8.16 branch -- pls re-verify it in a couple of hours (CI runs anew)

[zvonand]

03321_clickhouse_local_initialization_not_too_slow_even_under_sanitizers, test_dirty_pages_force_purge are both known and not important now, but @CarlosFelipeOR and @DimensionWieldr would give a better explanation here

[DimensionWieldr]

CI Failure Analysis (Post-Rebase to 25.8.16)

---

1. Regression Tests (aggregate_functions_1) - Pre-existing Failure

Failure: /aggregate functions/part 1/uniqTheta (both aarch64 and release)

Evidence these are pre-existing on base branch (releases/25.8.16):

• MasterCI #21704306488 (Feb 05): AggregateFunctions (1) = failure
• MasterCI #21640745419 (Feb 03): AggregateFunctions (1) = failure

Root cause: uniqTheta was updated in upstream.

Verdict: Pre-existing failure on base branch. Unrelated to PR changes.

---

2. Regression Tests (S3 suites) - Pre-existing Failures

Failures:

• s3_aws_s3_2 (aarch64 and release)
• s3_azure_2 (aarch64 and release)
• s3_minio_2 (aarch64 and release)

Evidence these are pre-existing on base branch:

• MasterCI #21704306488: All S3 (2) jobs = failure
• MasterCI #21640745419: All S3 (2) jobs = failure

Root cause: Zero copy replication tests failing - tracked in Altinity/clickhouse-regression#93

Verdict: Pre-existing failures on base branch. Unrelated to PR changes.

---

3. Infrastructure Failures (GitHub Actions Job Failures)

Failures:

• Integration tests (amd_asan, old analyzer, 1/6) - Job failed at 47m6s
• Integration tests (amd_asan, old analyzer, 2/6) - Job failed at 2h9m26s
• Stateless tests (amd_msan, parallel, 2/2) - Job failed at 47m6s

Evidence: These jobs failed at the GitHub Actions level without producing test results in the praktika report. The two 47m6s failures occurred at the same timestamp, suggesting infrastructure termination rather than test failures.

Verdict: CI infrastructure failures. Unrelated to PR changes.

---

4. Stateless Tests (amd_tsan, sequential, 2/2)

Failure: 03321_clickhouse_local_initialization_not_too_slow_even_under_sanitizers

What the test does: Measures if clickhouse-local can start 10 times in 10 seconds under sanitizers.

Code path analysis: The PR modifies merge selector and predicate code for TTL/cold volumes. This test only measures client startup speed - zero code path overlap with PR changes.

Verdict: Timing-sensitive performance test with no code path overlap. Unrelated to PR changes.

---

5. Integration Tests - test_dirty_pages_force_purge

Failures:

• Integration tests (amd_binary, 2/5) - fail: 1, passed: 1011
• Integration tests (arm_binary, distributed plan, 2/4) - fail: 1, passed: 988

What the test does: Tests jemalloc memory dirty page purge behavior via MemoryAllocatorPurge events.

Root cause: Test fails due to lower kernel version on CI runners.

Verdict: Infrastructure/environment issue. Unrelated to PR changes.

---

6. Regression Tests (iceberg_2)

Failure: data_types_check_where assertion failed - expected '0' but got empty result

Base branch status:

• MasterCI #21704306488: Iceberg (2) = success

Code path analysis:

• PR modifies: TTL merge selector, merge predicates for parts on cold volumes
• Test uses: Iceberg table engine data type handling with WHERE clauses
• No code path overlap - PR does not modify Iceberg engine or query execution

Verdict: Test passed on base branch but failed on PR. Code path analysis shows no overlap with PR changes. Likely flaky.

---

Summary

Test Category	Failure Type	Related to PR?
Regression aggregate_functions_1 (2 jobs)	Pre-existing (uniqTheta upstream change)	No
Regression S3 (6 jobs)	Pre-existing (zero copy replication)	No
ASAN/MSAN jobs (3 jobs)	Infrastructure failure	No
03321_clickhouse_local_init...	No code path overlap	No
test_dirty_pages_force_purge (2 jobs)	Lower kernel version on runners	No
Regression iceberg_2	No code path overlap	No

Verdict: All failures are unrelated to the PR changes.