Running Cortex on Kubernetes

[charannaik]

The rolling update strategy with maxSurge, maxUnavailable does not apply to the statefulsets and hence not applicable to ingesters.

[adeverteuil]

Link for context: https://cortexmetrics.io/docs/guides/running-cortex-on-kubernetes/#take-extra-care-with-ingesters

We're running GEM on Kubernetes with the enterprise-metrics Helm chart: https://github.com/grafana/helm-charts/blob/main/charts/enterprise-metrics/templates/ingester-statefulset.yaml

[stale[bot]]

This issue has been automatically marked as stale because it has not had any activity in the past 60 days. It will be closed in 15 days if no further activity occurs. Thank you for your contributions.

[charannaik]

Reopening. Need clarification on the reported.

[alanprot]

This seems odd indeed. Maybe we should to replace that with the PDB documentation.

https://github.com/cortexproject/cortex-helm-chart/blob/master/templates/ingester/ingester-poddisruptionbudget.yaml

PRs are welcome! :D

[kd7lxl]

Maybe we should to replace that with the PDB documentation.

PDBs and update strategies apply to different cases. PDBs only apply in the context of node maintenance, like draining a node. The rollout policy applies to rolling out changes to the pod spec. For high availability, both policies must be in place. It makes sense for them to have corresponding policies, like each specifying maxUnavailable: 1.

[alvinlin123]

Reopening. Need clarification on the reported.

I just learned that the document in question was written assuming Deployment. That's why the doc talks about maxUnavailable and maxSurge for ingesters.

[stale[bot]]

This issue has been automatically marked as stale because it has not had any activity in the past 60 days. It will be closed in 15 days if no further activity occurs. Thank you for your contributions.