Issues

Issues

• PREWHERE profile events count ORDER BY ... LIMIT top-K filter rejections

  ClickHouse/ClickHouse#131041 · opened on Nov 12, 2026

• Buffer flush ignores KILL in the Distributed delay; restart loses rows

  ClickHouse/ClickHouse#131024 · opened on Nov 12, 2026
  2

• Text index merge uses stale _part_index during pending column rename

  ClickHouse/ClickHouse#131010 · opened on Nov 12, 2026
  2

• Bare tables in FLUSH ASYNC INSERT QUEUE ON CLUSTER resolve in host default DB

  ClickHouse/ClickHouse#131005 · opened on Nov 12, 2026
  1

• a.b of a StripeLog Merge child still 0 with an ALIAS or remote sibling

  ClickHouse/ClickHouse#131000 · opened on Nov 12, 2026

• Row policy on a StripeLog child's t.a still fails through merge

  ClickHouse/ClickHouse#130998 · opened on Nov 12, 2026

• PromQL misses external samples after TRUNCATE with inner metric families

  ClickHouse/ClickHouse#130994 · opened on Nov 12, 2026
  2

• TRUNCATE ALL TABLES fails on external TimeSeries, re-inserted series vanish

  ClickHouse/ClickHouse#130993 · opened on Nov 12, 2026

• abfss storage_endpoint gets its container appended twice for Iceberg tables

  ClickHouse/ClickHouse#130989 · opened on Nov 12, 2026

• Ordered S3Queue stalls after buckets 1 → 0 → 2 with legacy bucket-0 pointer

  ClickHouse/ClickHouse#130987 · opened on Nov 12, 2026

• Flaky test: 04105_limit_by_into_sort_explain

  ClickHouse/ClickHouse#130985 · opened on Nov 12, 2026

• *Cluster outer filter hits wrong column when ORDER BY key is not selected

  ClickHouse/ClickHouse#130980 · opened on Nov 12, 2026

• JSON DISTINCT/uniqExact merge distinct objects that GROUP BY keeps apart

  ClickHouse/ClickHouse#130970 · opened on Nov 12, 2026

• Parquet columns cache: rows read by lazy materialization are never cached; caching them under the set of requested rows makes repeated SELECT * ... ORDER BY ... LIMIT 60-89% faster (ClickBench Q23: 82 -> 33 ms)

  ClickHouse/ClickHouse#130969 · opened on Nov 12, 2026

• S3 BACKUP/RESTORE with zero initial retry backoff aborts debug builds

  ClickHouse/ClickHouse#130968 · opened on Nov 12, 2026

• Request: Backport #119760 to 26.8

  ClickHouse/ClickHouse#130951 · opened on Nov 12, 2026
  1

• Parquet lazy materialization reads whole column chunks when there is no offset index; locating pages from their headers makes ClickBench Q23 7-15% faster (new server per query) and SELECT * ... LIMIT 100-1000 19-32% faster

  ClickHouse/ClickHouse#130939 · opened on Nov 11, 2026

• Azure blob reads can open hundreds of connections for one small query

  ClickHouse/ClickHouse#130936 · opened on Nov 11, 2026
  1

• (NOT x) | f in call-form contexts formats back as NOT (x | f)

  ClickHouse/ClickHouse#130934 · opened on Nov 11, 2026

• A re-attached table loads parts from a plain_rewritable listing that nothing refreshes

  ClickHouse/ClickHouse#130933 · opened on Nov 11, 2026
  1

• Iceberg performance on TPC-DS and TPC-H with use_iceberg_manifest_statistics and use_iceberg_table_statistics_ndv

  ClickHouse/ClickHouse#130932 · opened on Nov 11, 2026

• Iceberg forces in-order Parquet reading on every data file; limiting it to files with position deletes makes queries up to 8% faster

  ClickHouse/ClickHouse#130931 · opened on Nov 11, 2026

• if over String columns copies every row separately; copying each run of rows that take the same branch at once makes ClickBench Q39 over 100 Parquet files 5-7% faster (MergeTree 3%)

  ClickHouse/ClickHouse#130929 · opened on Nov 11, 2026

• TimeSeries target named by {t:Identifier} with UUID aborts debug server

  ClickHouse/ClickHouse#130928 · opened on Nov 11, 2026
  1

• Parquet columns cache: columns read after PREWHERE has dropped rows of their chunk are not cached; caching them under the set of rows that passed makes repeated filtered ClickBench queries 14-63% faster

  ClickHouse/ClickHouse#130925 · opened on Nov 11, 2026