Skip to content

Latest commit

 

History

History
33 lines (20 loc) · 1.59 KB

README.md

File metadata and controls

33 lines (20 loc) · 1.59 KB

multi-k8s-crdb

A helm chart to span a cockroachdb instance across k8s clusters

Attribution

This chart is based heavily on the official CockroachDB chart. This chart is meant to include additional templates Equinix Metal relies on.

Memory allocation

This chart sets max-sql-memory and cache to 1GB, which is 25% of the resource memory limit. If you change the memory limit, you can adjust those two parameters accordingly.

A note on upgrades

When doing a major version upgrade, you need to be on the latest minor version before that, e.g. to upgrade to v22.1 you need to be running v21.2.x. In addition, to prevent the cluster from automatically finalizing when all the cluster nodes are upgraded to the new major version, you should set this before you start the upgrade:

SET CLUSTER SETTING cluster.preserve_downgrade_option = '21.2'

It will allow you to downgrade the cluster if you run into any issues. You can run in this state for a few days and if everything looks good, finalize the upgrade:

RESET CLUSTER SETTING cluster.preserve_downgrade_option

For more information about upgrades, see https://www.cockroachlabs.com/docs/stable/upgrade-cockroachdb-kubernetes.html

Breaking changes

  • This chart does not support the Cockroach Labs self signer.
  • This chart assumes there will be externally routable pod endpoints, with dns provided by ExternalDNS
  • Assumes the use of Stakater's Reloader for cert rotation.