Issues

• Shredded variant COW upsert fails when the base file stores the variant as REQUIRED

  priority:hightype:bug
  apache/hudi#20349 · Filter by author voonhous opened on Oct 15, 2026
  

• [BUG] Downgrade below version 9 records several ordering fields as a single precombine field

  type:bug
  apache/hudi#20347 · Filter by author wombatu-kun opened on Oct 15, 2026

• [BUG] Tables below version 9 lack hoodie.table.precombine.field, so 0.x and 1.0.x see no ordering field

  type:bug
  apache/hudi#20345 · Filter by author wombatu-kun opened on Oct 15, 2026

• [SUPPORT] Spark FGR: sampled size extrapolation triggers spill for long-tailed UnsafeRows

  type:community-support
  apache/hudi#20344 · Filter by author q2788116586-tech opened on Oct 15, 2026

• HoodieIndexer rebuilds a partitioned record-level index as a global one

  apache/hudi#20343 · Filter by author nsivabalan opened on Oct 15, 2026

• [Flink] Speed up streaming split discovery by batching file existence checks per directory

  type:devtask
  apache/hudi#20338 · Filter by author skywalker0618 opened on Oct 14, 2026
  3

• [FEATURE] Recover table when hoodie.properties is corrupted or missing

  apache/hudi#20335 · Filter by author y2ksayed opened on Oct 14, 2026
  

• Flink global failover can trigger concurrent commits for the same instant

  apache/hudi#20331 · Filter by author cshuo opened on Oct 14, 2026

• Preserve delete payloads in Spark record reconstruction

  apache/hudi#20330 · Filter by author cshuo opened on Oct 14, 2026

• [BUG] Log compaction rewrites record commit times; incremental reads re-emit old rows after cleaning (1.2.1)

  type:bug
  apache/hudi#20327 · Filter by author prakashgyank46-commits opened on Oct 14, 2026

• [BUG] Log compaction scheduling throws NPE when CDC log blocks are present (Hudi 1.2.1)

  type:bug
  apache/hudi#20325 · Filter by author prakashgyank46-commits opened on Oct 14, 2026
  1

• Variant type certification: per-type storage matrix across shredded layouts and record types

  apache/hudi#20322 · Filter by author voonhous opened on Oct 14, 2026
  1

  

• [BUG] flink: bucket pruning misses rows on tables with one primary key and multiple partition fields

  engine:flinktype:bug
  apache/hudi#20313 · Filter by author wombatu-kun opened on Oct 14, 2026

• [Improvement] Avoid cross-partition hot-bucket concentration in RemotePartitioner

  type:devtask
  apache/hudi#20306 · Filter by author K1taIkuyo0421 opened on Oct 13, 2026
  3

• Spark data skipping drops matching rows after adding columns to the column stats index

  engine:sparktype:bug
  apache/hudi#20296 · Filter by author cshuo opened on Oct 13, 2026

• Flink timestamp predicates lose microsecond precision during filtering and index lookup

  engine:flinktype:bug
  apache/hudi#20293 · Filter by author cshuo opened on Oct 13, 2026

• Flink bucket pruning misses compound keys containing placeholder literals

  engine:flinktype:bug
  apache/hudi#20291 · Filter by author cshuo opened on Oct 13, 2026

• Support vectorized Parquet base reads in Spark MOR FileGroupReader

  type:devtask
  apache/hudi#20277 · Filter by author q2788116586-tech opened on Oct 10, 2026

• Tables created below version 8 omit hoodie.table.keygenerator.class

  area:corepriority:mediumtype:bug
  apache/hudi#20275 · Filter by author yihua opened on Oct 9, 2026

• Partial-update merging compares Avro schemas field by field for every record

  area:performancearea:readerpriority:mediumtype:improvement
  apache/hudi#20268 · Filter by author yihua opened on Oct 9, 2026
  1

• Data skipping returns too few rows on a METADATA_ONLY bootstrapped table once column stats are enabled

  area:metadata-tablearea:readercomponent:bootstrapcomponent:data-skippingengine:sparkpriority:hightype:bug
  apache/hudi#20266 · Filter by author yihua opened on Oct 9, 2026

• Timeline server serializes every file system view request on the view lock while hashing the timeline

  area:performancearea:timelinepriority:hightype:devtask
  apache/hudi#20264 · Filter by author yihua opened on Oct 9, 2026

• Bloom index key check asks the timeline server for each candidate base file twice per task

  area:indexarea:performancepriority:mediumtype:devtask
  apache/hudi#20262 · Filter by author yihua opened on Oct 9, 2026

• MOR upserts silently drop records when hoodie.allow.operation.metadata.field is enabled

  area:writerpriority:hightype:bug
  apache/hudi#20260 · Filter by author yihua opened on Oct 9, 2026

• Read each metadata table native HFile log file once instead of three times

  area:metadata-tablearea:performancepriority:hightype:devtask
  apache/hudi#20258 · Filter by author yihua opened on Oct 9, 2026