Issues

• --recovery-target-* flags are parsed but ignored by the server binary (point-in-time recovery is library-only)

  pilotspace/moon#1327 · Filter by author TinDang97 opened on Oct 2, 2026

• --migrate-aof-* routes every record by its first argument: multi-key commands, XGROUP and SELECT n>0 produce a wrong or partial per-shard dataset

  pilotspace/moon#1326 · Filter by author TinDang97 opened on Oct 2, 2026

• AOF: a MULTI/EXEC body has no block markers, so a crash mid-body replays a partial transaction (redis wraps it in MULTI/EXEC and reverts it)

  pilotspace/moon#1325 · Filter by author TinDang97 opened on Oct 2, 2026
  1

• monoio: SCAN omits spilled (cold) keys — returns ~90% of the keyspace while DBSIZE and GET are intact

  pilotspace/moon#1324 · Filter by author TinDang97 opened on Oct 2, 2026

• Admin console gateway: writes under appendfsync always are acked without an AOF fsync barrier

  pilotspace/moon#1323 · Filter by author TinDang97 opened on Oct 2, 2026

• appendfsync always: cross-shard MSET/DEL/UNLINK (and other multi-key coordinators) ack before the remote shard's AOF fsync

  pilotspace/moon#1322 · Filter by author TinDang97 opened on Oct 2, 2026

• A single-file appendonly.aof booted with --shards N>1 is replayed into every shard (DBSIZE ×N); recovery ignores --appendfilename

  pilotspace/moon#1321 · Filter by author TinDang97 opened on Oct 2, 2026

• Replica can make a never-committed TXN permanent after a master restart + partial resync

  pilotspace/moon#1320 · Filter by author TinDang97 opened on Oct 1, 2026

• BGSAVE at --shards > 1 is not point-in-time: a write acknowledged after "+Background saving started" can land in the snapshot

  pilotspace/moon#1319 · Filter by author TinDang97 opened on Oct 1, 2026

• A replica with appendonly yes does not log its master's stream to its own AOF: restarted on its own it comes back empty

  pilotspace/moon#1318 · Filter by author TinDang97 opened on Oct 1, 2026
  1

• Replica full sync drops per-field hash TTLs (HPTTL on the replica reads -1)

  pilotspace/moon#1317 · Filter by author TinDang97 opened on Oct 1, 2026

• Runtime divergence: a tokio GET of a cold key can serve it without promoting it (monoio always promotes), so cold keys, their spill files and later read latency differ by runtime

  pilotspace/moon#1315 · Filter by author TinDang97 opened on Oct 1, 2026

• After the AOF writer latches a write error (EFBIG/EDQUOT/EIO), appendfsync everysec keeps acking writes, drops them from the AOF, and INFO still says aof_last_write_status:ok

  pilotspace/moon#1314 · Filter by author TinDang97 opened on Sep 30, 2026
  1

• Replica judges key expiry by its own clock on master-stream lookups: a lagging replica misses PERSIST/PEXPIRE on a key its master kept (reads nil where redis's replica reads the value)

  pilotspace/moon#1313 · Filter by author TinDang97 opened on Sep 30, 2026

• CLIENT KILL of a client with an open TXN: the killer's +OK arrives before the victim's rollback, so the victim's keys still answer -TXNCONFLICT under appendfsync always

  pilotspace/moon#1312 · Filter by author TinDang97 opened on Sep 30, 2026

• TXN key holds (moon#1299 follow-up): over-refusal of LMPOP/option-less GETEX/GET-only BITFIELD, and a cross-shard BLPOP blocks where a same-shard one is refused

  pilotspace/moon#1311 · Filter by author TinDang97 opened on Sep 30, 2026

• Redis 7.2 parity: ACL SETUSER accepts rules redis rejects, key-pattern merging, LIST/SAVE omit sanitize-payload/resetchannels; XADD MAXLEN abc accepted

  pilotspace/moon#1310 · Filter by author TinDang97 opened on Sep 30, 2026

• appendfsync everysec: an fsync error never refuses writes (redis answers -MISCONF until a later fsync succeeds)

  pilotspace/moon#1309 · Filter by author TinDang97 opened on Sep 30, 2026

• REPLICAOF on a node with an open TXN: TXN COMMIT/ABORT answer READONLY, and the disconnect abort restores pre-images onto the replica (diverges from its master)

  pilotspace/moon#1308 · Filter by author TinDang97 opened on Sep 30, 2026
  1

• TXN isolation covers KV keys only: TXN.ABORT restores a graph property over another client's acknowledged graph write

  pilotspace/moon#1307 · Filter by author TinDang97 opened on Sep 30, 2026

• ACL GETUSER/LIST/SAVE expand category rules (+@read) into commands instead of keeping the token (redis parity)

  pilotspace/moon#1306 · Filter by author TinDang97 opened on Sep 30, 2026

• TXN: a connection write that answers an error keeps its undo capture, so TXN.ABORT can restore over another client's write

  pilotspace/moon#1303 · Filter by author TinDang97 opened on Sep 28, 2026

• Graph writes past the 4096-slot WAL append channel are acked but not persisted (kill -9 loses them)

  pilotspace/moon#1302 · Filter by author TinDang97 opened on Sep 28, 2026
  1

• TXN crash atomicity: kill -9 inside an open TXN keeps its uncommitted writes (AOF); a BGSAVE during a TXN persists them (no AOF)

  bug
  pilotspace/moon#1300 · Filter by author TinDang97 opened on Sep 28, 2026
  1

• TXN isolation: another client's write to a key held by an open TXN is silently overwritten by the TXN's ABORT (block other writers)

  bug
  pilotspace/moon#1299 · Filter by author TinDang97 opened on Sep 28, 2026
  1