Issues

• IngestExternalFiles with atomic_replace_range misses a memtable range tombstone when another tombstone fragment starts at the range limit: no flush, the ingested file goes below the older tombstones and its keys are lost (call returns OK)

  facebook/rocksdb#15440 · Filter by author hanke580 opened on Oct 20, 2026

• format_version=8 with ReverseBytewiseComparator: a data block holding k + "\x01" before k stores a common prefix longer than k, so Get/MultiGet/Seek of k skip its newest version (stale value, deleted value back, or NotFound)

  facebook/rocksdb#15439 · Filter by author hanke580 opened on Oct 20, 2026

• Process-lifetime lock leak when DB close is skipped: locked_files entry is never removed

  facebook/rocksdb#15432 · Filter by author RanjaniRB opened on Oct 19, 2026

• WritePrepared/WriteUnprepared: rolling back a prepared transaction strips all non-default columns of a wide-column entity it touched

  facebook/rocksdb#15426 · Filter by author hanke580 opened on Oct 18, 2026

• cf_allow_ingest_behind: compaction still drops range tombstones as obsolete, so an ingest_behind back-fill resurrects DeleteRange'd keys (point deletes are preserved since #13807)

  facebook/rocksdb#15425 · Filter by author hanke580 opened on Oct 18, 2026

• User-defined timestamps + Merge: compaction with full_history_ts_low merges the newest version below the cutoff into a newer operand, reads at timestamps >= cutoff return NotFound

  facebook/rocksdb#15424 · Filter by author hanke580 opened on Oct 18, 2026

• enable_blob_files + preclude_last_level_data_seconds: compaction writes blob files for proximal-level outputs but never registers them ("Corruption: Invalid blob file number", values lost permanently)

  facebook/rocksdb#15423 · Filter by author hanke580 opened on Oct 18, 2026

• RepairDB with recycle_log_file_num > 0 replays the recycled (already flushed) WAL: merge operands are applied twice

  facebook/rocksdb#15422 · Filter by author hanke580 opened on Oct 18, 2026

• RepairDB undoes IngestExternalFile on an intact DB: files from SstFileWriter without a CF handle are moved to lost/, and ingested files lose their global seqno (older values win again)

  facebook/rocksdb#15421 · Filter by author hanke580 opened on Oct 18, 2026

• Compaction merges across a snapshot at sequence number 0 (e.g. taken right after ingesting into a new DB): DeleteRange'd data comes back permanently and the snapshot loses its value

  facebook/rocksdb#15420 · Filter by author hanke580 opened on Oct 18, 2026

• Get() drops the base value of a merge when the key's versions span two SST files of one level and the first file has an older covering range tombstone (MultiGet and iterators are correct)

  facebook/rocksdb#15419 · Filter by author hanke580 opened on Oct 18, 2026

• Null dereference in WriteThread::EnterAsBatchGroupLeader when the r_list graft's compare_exchange_weak fails spuriously (aarch64 LL/SC)

  facebook/rocksdb#15414 · Filter by author SebastianThiebaud opened on Oct 16, 2026
  1

• Java release for 11.12

  facebook/rocksdb#15385 · Filter by author meta-rocksdb-release-automation opened on Oct 12, 2026
  

• RocksJava: expose explicit-length byte[] overloads for write-batch put, merge and delete

  facebook/rocksdb#15381 · Filter by author evgenydalakyan opened on Oct 11, 2026

• Iterator returns freed memory when a merge operator returns a blob base value via existing_operand (e.g. MaxOperator)

  facebook/rocksdb#15336 · Filter by author ronag opened on Oct 6, 2026

• BlockBasedTableOptions::super_block_alignment_size accepts values at or below block_size, which its documentation says are invalid

  facebook/rocksdb#15318 · Filter by author hanke580 opened on Oct 4, 2026
  1

• DB::Open with a negative num_levels terminates the process (uncaught std::length_error in MutableCFOptions::RefreshDerivedOptions) instead of returning a Status

  facebook/rocksdb#15317 · Filter by author hanke580 opened on Oct 4, 2026
  1

• TransactionDB::SetDeadlockInfoBufferSize(n) with n == current number of recorded deadlocks leaves buffer_idx_ == size(); next deadlock writes past the end of paths_buffer_

  facebook/rocksdb#15316 · Filter by author hanke580 opened on Oct 4, 2026
  2

• Leveled compaction: ttl greater than the current unix time underflows current_time - ttl, so every file is treated as TTL-expired

  facebook/rocksdb#15315 · Filter by author hanke580 opened on Oct 4, 2026

• WriteBatch entry count silently wraps at 2^32 (Put/Delete return OK); the batch then fails DB::Write with "Corruption: WriteBatch has wrong count" and stops the DB

  facebook/rocksdb#15314 · Filter by author hanke580 opened on Oct 4, 2026

• Flush() returns OK but writes a >= 4 GiB data block that the block reader truncates; every key in the block becomes unreadable

  facebook/rocksdb#15313 · Filter by author hanke580 opened on Oct 4, 2026
  3

• Blob GC improvement for better space and write amp

  facebook/rocksdb#15301 · Filter by author xingbowang opened on Oct 2, 2026

• Some remaining costly misses in rocksjni

  facebook/rocksdb#15295 · Filter by author alanpaxton opened on Sep 30, 2026

• db_bench: --report_file keeps only the last benchmark when --benchmarks lists several

  facebook/rocksdb#15276 · Filter by author vinibarbieri opened on Sep 27, 2026

• Java release for 11.11

  facebook/rocksdb#15266 · Filter by author meta-rocksdb-release-automation opened on Sep 24, 2026