Restarting all data nodes

I currently have a cluster that has dedicated data nodes that use PVC for storage. I am testing various configurations and as it stands now when I make a change using the manifest, the controller will take down each node one-by-one, keeping the cluster in a yellow / green state. However, due to the high number of nodes (and data stored within) this will take up to a day.

I'm curious if I can make the configuration change and then simply delete each of the data pods explicitly, accepting the fact that the cluster will enter a red state temporarily but eventually come back up and re-sync. This will be a much faster operation overall as the nodes can each come back up in parallel, accepting the fact that the state will go red until it is resync'd.

You can increase the maxUnavailable setting to tune this behavior:

Though manually killing pods will also work, it is not the first lever I would reach for.

Great -- thanks for the link / info!