cockroachdb / cockroachdb/cockroach
rangefeed: consider rangefeed resource usage and load balancing
Open
A-kv-rangefeed
A-kv-replication
C-enhancement
T-kv
- Dominant language
- Go
- Stars
- 32.5k
- Forks
- 4.1k
- PR merge metrics
- PR metrics pending
Description
Currently rangefeeds are created on the closest replica to the client.
I some node serves lots of rangefeeds for idle ranges this overhead will be completely under the radar and uncorrelated to other read/write activity. It could get particularly bad if we do rolling restart and lots of processors accumulate on a single node.
We need some way to account for rangefeed resource usage to measure the impact in real deployments and if it becomes problematic also means to shed processors off node if they cause excessive load.
Jira issue: CRDB-29426
Epic CRDB-37617
Contributor guide
Assessment
This issue has not been assessed yet.