cockroachdb / cockroachdb/cockroach
Provide a better sampling strategy by statement fingerprint for query telemetry data
- Dominant language
- Go
- Stars
- 32.5k
- Forks
- 4.1k
- PR merge metrics
- PR metrics pending
Description
The current sampling strategy for our telemetry logging channel for the query log is too aggressive where we can drop and miss critical events for a workload. We wait 1/10 (default setting) of a second before sampling the next query across the entire cluster. This can cause us to misunderstand our customer’s application development lifecycle and database feature adoption. We should improve the sampling strategy for our query telemetry data. We should investigate and improve our sampling strategy by sampling by statement fingerprint.
This will help with _several_ initiatives
- End users externalizing and consuming events from the telemetry logging channel can have finer control and granularity on the information they are capturing.
- Internal users working on [critical data initiatives ](https://cockroachlabs.slack.com/archives/CL94XLGG1/p1676055419565399?thread_ts=1676048412.721649&cid=CL94XLGG1) have a more holistic picture of workload including steps in their onboarding journey
Jira issue: CRDB-28052
Epic CRDB-32141
Contributor guide
Assessment
This issue has not been assessed yet.