cockroachdb / cockroachdb/cockroach
kv,obs: persist lease history beyond 5 changes in the rangelog contained in debug.zip
- Dominant language
- Go
- Stars
- 32.5k
- Forks
- 4.1k
- PR merge metrics
- PR metrics pending
Description
Very often, operator's are tasked with identifying the cause for an issue in the past. Knowing where particular ranges held their lease is very valuable. For example, missing timeseries metrics can be explained by a network issue on the leaseholder for the timeseries range. Another example is a cluster struggling because the lease for the liveness table was held on a bad node.
Today, we persist 5 entries for the history in our logs, and those 5 entries can constantly change and we have no historical record to answer this. The following [thread](https://cockroachlabs.slack.com/archives/C0KB9Q03D/p1715617685630799) (internal) explains why it is 5 today, and an explanation on the impact if we were to increase it.
Jira issue: CRDB-38799
Contributor guide
Research direction
Start by locating the lease-history and rangelog implementation that produces debug.zip, then read how the current five-entry history is persisted and replaced. Trace the related observability or KV tests if present; done means debug.zip retains lease history beyond five changes without unacceptable impact described in the issue.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- go
- Domain
- databases, distributed-systems, observability
- Issue type
- Feature
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 35/100