glorv
glorv
/cc @rleungx @nolouch is this expected behavior ?
@nolouch The cluster is small(less than 10GB) and there is no workload. You can see full metrics here: https://clinic.pingcap.com/portal/#/orgs/1372813089196911825/clusters/10123226947374817763 The time range is: 2024.05.22 16:30~19:15
Tracking issue for the potential OOM cause by this feature: #17055
An intuitive idea: What about move write task batch to raft fsm side. Instead of send a write message after handle each region, we can first cache region write message...
> How about existing clusters already used rocksdb log configs upgrade to the new version? @zhangjinpeng87 The behavior is not changed, that is, from v5.0, rocksdb log config won't take...
@cfzjywxk PTAL