Open ezbz opened 3 years ago
Is there a possible workaround until the operation is supported by the operator? Note that I edited the statefulset directly but the operator eventually reverted.
Ignore the above. The Operator reverted my image tag updates to the statefulset, but my CPU & memory limits stuck. I have a simple workaround of modifying the statefulset directly for requests & limits, while using the operator to manage the image version.
yep, operator won't change the statefulset unless you edit/drop scylla-operator.scylladb.com/managed-hash
annotation or it actually needs to make an update
support for changing cpu and memory went in https://github.com/scylladb/scylla-operator/pull/763, storage can be resized manually on the PVC but we can't change the initial claim because StatefulSet forbids mutating it.
I just hit this issue when trying to resize the volumes. The documentation in https://github.com/scylladb/scylla-operator/blame/master/docs/source/scylla_cluster_crd.md#L169 should be fixed:
"Capacity of the PersistentVolume to request." -> "Initial capacity of the PersistentVolume to request."
It isn't obvious that you have to modify the PVCs to get an increased capacity. The operator will not try and reconcile the difference.
I guess we could make the wording less ambiguous if it managed to confuse someone. The request part is what should've implied the meaning - you make a "request" to get a PVC with size that's at least this big; it's just a size of the "request", not the size of the PVC. This is in line with how Kubernetes StatefulSets handle it, unfortunately the StatefulSet controller doesn't reconcile nor own the PVCs and we use StefulSets to represent and manage our racks.
yep, operator won't change the statefulset unless you edit/drop
scylla-operator.scylladb.com/managed-hash
annotation or it actually needs to make an updatesupport for changing cpu and memory went in #763, storage can be resized manually on the PVC but we can't change the initial claim because StatefulSet forbids mutating it.
@tnozicka do you know if there is any workaround for this issue?
Thanks,
Diego
There is a preview of the docs at https://github.com/scylladb/scylla-operator/pull/1053 (needs a review but could give you some pointers). As with any preview, test this on a toy cluster first.
The Scylla Operator project currently lacks enough contributors to adequately respond to all issues.
This bot triages un-triaged issues according to the following rules:
lifecycle/stale
is appliedlifecycle/stale
was applied, lifecycle/rotten
is appliedlifecycle/rotten
was applied, the issue is closedYou can:
/remove-lifecycle stale
/close
/lifecycle stale
Extending disk is partially supported, CPU and Mem are not.
Requested by Customers