Closed kkimdev closed 5 years ago
@kkimdev I don't know what you mean by nodes
? TiDB cluster pods or just k8s work nodes?
For scaling in TiDB cluster pods, it is as simple as changing the component replicas and run helm upgrade
as documented in the operation guide.
As for removing k8s worker nodes, there are some extra steps. We'll add the document soon. But if you are familiar with k8s, then after the component replica count is decreased, removing the k8s node can simply be achieved by kubectl drain node
command.
The current operation guide doesn't explain how to safely downscale TIDB. I suspect that it's not as simple as just draining the nodes to remove as it the data migration can take more than the default pod termination grace period(30sec) https://kubernetes.io/docs/concepts/workloads/pods/pod/#termination-of-pods . But in any case, it will be great to have it documented.