mittalrishabh comments

Results 49 comments of


                                            mittalrishabh

raftstore: accumulated read index requests may result in a large number of re-propose actions, increasing pressure on the store loop

1. Can we batch the read index if async commit is disabled ? 2. Can we send memory locks along with the writes through raft to the follower. I strongly...

raftstore: accumulated read index requests may result in a large number of re-propose actions, increasing pressure on the store loop

>>It's checking memory locks on the leader peer when receiving ReadIndexRequest to make sure the snapshot isolation >>is guaranteed, but not sending memory locks. Can't we maintain another copy of...

TiKV should be tolerant on single EBS IO jitter

Have you decided the release for this feature ?

TiKV should be tolerant on single EBS IO jitter

Can you elaborate on features those have dependency on data being persisted in raft disk. Is it CDC ?

TiKV should be tolerant on single EBS IO jitter

There are two challenges associated with raft disk jitter: 1. Write latency increases, followed by an increase in the error rate. 2. Follower read latency can increase if a replica...

[DNM] Wal debug 6.29.tikv

@tonyxuqqi What is the WAL recovery mode in tikv ? I guess that it should be kTolerateCorruptedTailRecords.

which format is tikv using. Seems like rocksdb wal has two formats : legacy and recycle. Recycle format has more debug information (log number) . https://github.com/facebook/rocksdb/wiki/Write-Ahead-Log-File-Format

raftstore: accumulated read index requests may result in a large number of re-propose actions, increasing pressure on the store loop

Thanks. I will go through the above link. Does it make sense to send batch read index to reduce the number of RPC calls.

split_controller : add logging for hot keys and region split key

I think it needs one more review

split_controller : add logging for hot keys and region split key

/retest