cockroachdb / cockroachdb/cockroach
kvserver: add cross-az and cross-region replication latency
- Dominant language
- Go
- Stars
- 32.5k
- Forks
- 4.1k
- PR merge metrics
- PR metrics pending
Description
**Is your feature request related to a problem? Please describe.**
We added replication latencies in https://github.com/cockroachdb/cockroach/pull/106094, but they're not aware of topology.
**Describe the solution you'd like**
Recently, @wenyihu6 added machinery to track topology for metrics: https://github.com/cockroachdb/cockroach/pull/105122
We should adopt the same approach for replication latency. Concretely, when a proposer's config indicates that a quorum is cross-region _or_ that the raft leader is cross-region, we should classify as cross-region, etc.
This is a bit more involved than the existing cross-region metrics and will likely involve updating some state whenever the raft leader or replication config changes, as computing this state for each proposal is likely prohibitively expensive.
**Describe alternatives you've considered**
**Additional context**
Replication latency is a great candidate for becoming an SLI we care about, so they should be powered sufficiently to get the kind of observability we need. Letting cross-region replication "crowd out" replication between local followers is not ideal.
Jira issue: CRDB-29591
Contributor guide
Assessment
This issue has not been assessed yet.