cockroachdb / cockroachdb/cockroach
kvserver: range availability metrics should incorporate Raft state
- Dominant language
- Go
- Stars
- 32.5k
- Forks
- 4.1k
- PR merge metrics
- PR metrics pending
Description
Range availability metrics, e.g. `ranges.underreplicated` and `ranges.unavailable`, are calculated solely based on the node liveness state. This can be unreliable in cases where nodes are able to heartbeat, but replicas/ranges are unavailable from the point of view of the Raft leader or leaseholder, for example under partial/asymmetric network partitions, disk stalls, or other failure modes.
These checks should additionally take into account e.g. the Raft state, lease state, and replica circuit breakers. See also #61118.
https://github.com/cockroachdb/cockroach/blob/ac3e633279c7333fe6544d1376593f83f57a5add/pkg/kv/kvserver/replica_metrics.go#L215-L235
Jira issue: CRDB-27677
Contributor guide
Assessment
This issue has not been assessed yet.