- Define SLOs for the "opensearch-as-a-service" / opensearch on k8s project
- Create SLO documentation on wikitech based on these instructions
Description
Description
Details
Details
- Other Assignee
- RKemper
| Status | Subtype | Assigned | Task | ||
|---|---|---|---|---|---|
| Open | None | T362105 EPIC: OpenSearch on K8s (formerly Mutualized opensearch cluster) - FY25/26 WE4.2.6 | |||
| Duplicate | bking | T416167 Create SLOs for OpenSearch on k8s | |||
| Duplicate | bking | T416269 Create SLIs for opensearch on k8s |
Event Timeline
Comment Actions
(Discussed with bking, brief summary follows)
SLIs to look at:
- Uptime of k8s cluster
- Availability of Opensearch API
- Availability of opensearch operator (i.e. it can still manage the cluster properly)
Corresponding SLOs:
- Uptime of k8s cluster => 95%
- Availability of Opensearch => 95%
- Availability of opensearch operator => 95%
Other metrics like indexing latency, query latency, etc aren't good candidates for SLOs, because of how free-form this service is. Users are free to choose pretty much any schema, so we can't make guarantees about the performance of the cluster under arbitrary datasets. These do seem like good metrics to be used for normal monitoring & alerting, however.