Joshua Moody

@chris93111 if you care about the filesystem size in the container, you should be monitoring the kubelet volume metrics, ref: https://longhorn.io/docs/1.2.4/monitoring/kubelet-volume-metrics/#about-kubelet-volume-metrics

If you allocate 5GB for your volume for example and the volume currently has an actual size of 4.9GB, it shouldn't matter whether the filesystem consumes only 300mb of that...

@kirinnee Currently we don't support trim, so the only advice I can give don't over provision your PVC, allocate the maximum expected file system size as your PVC volume size....

@kirinnee can you flip the switch to `show System Hidden` I am guessing you had some replica rebuild events, that left system snapshots behind?

Not at the moment, best advice is to not over provision your volume during creation. If you are only expecting 10GB usage then no need to provision 100GB, since you...

@derekbit good job on the evaluation :)

@w13915984028 please share the default settings .yaml/helm file that you used. As well as the current settings, you can get us the current settings by just uploading a support bundle.

ref for PDB issue https://github.com/longhorn/longhorn/issues/4689#issuecomment-1271285060

Here is an old PR where I explored the VA refactoring, this wasn't using a dedicated VA crd instead there was dedicated controllers per operation that would be managing desired...

@derekbit I agree that it would be simpler to start without active/passive (i.e. no second engine) and just rely on the share-manager pod being started on a different node. Grace...