feature: support reschedule disk with resource exhausted error in del… #890
Conversation
k8s-ci-robot
added
kind/feature
|
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: mowangdk The full list of commands accepted by this bot can be found here. The pull request process is described here
Needs approval from an approver in each of these files:
Approvers can indicate their approval by writing |
k8s-ci-robot
added
size/XS
mowangdk
force-pushed
the
feature/support_reschedule_pvc_prebind
branch
from
6e3a796 to
ee8b39d
Compare
| @@ -969,7 +969,7 @@ func getDiskType(diskVol *diskVolumeArgs) ([]string, []string, error) { | |||
| nodeInfo, err := client.CoreV1().Nodes().Get(context.Background(), diskVol.NodeSelected, metav1.GetOptions{}) | |||
| if err != nil { | |||
| log.Log.Infof("getDiskType: failed to get node labels: %v", err) | |||
| goto cusDiskType | |||
| return nil, nil, status.Errorf(codes.ResourceExhausted, "CreateVolume:: get node info by name: %s failed with err: %v, start to reschedule", diskVol.NodeSelected, err) | |||
There was a problem hiding this comment.
How about just return ResourceExhausted if we got a NotFound err. And return Internal error for other err. To avoid possible infinite rescheduling loop.
There was a problem hiding this comment.
The retry time follows the backoff algorithm and has little effect on CSI. But it still has side effect on the scheduler, so I'll modify it.
| provisionDiskTypes := []string{} | ||
| allTypes := deleteEmpty(strings.Split(diskVol.Type, ",")) | ||
| if len(nodeSupportDiskType) != 0 { | ||
| provisionDiskTypes = intersect(nodeSupportDiskType, allTypes) | ||
| if len(provisionDiskTypes) == 0 { | ||
| log.Log.Errorf("CreateVolume:: node(%s) support type: [%v] is incompatible with provision disk type: [%s]", diskVol.NodeSelected, nodeSupportDiskType, allTypes) | ||
| return nil, nil, status.Errorf(codes.InvalidArgument, "CreateVolume:: node support type: [%v] is incompatible with provision disk type: [%s]", nodeSupportDiskType, allTypes) | ||
| return nil, nil, status.Errorf(codes.ResourceExhausted, "CreateVolume:: node support type: [%v] is incompatible with provision disk type: [%s]", nodeSupportDiskType, allTypes) |
There was a problem hiding this comment.
We can only reach this line if the CSI plugin is just installed, and scheduling is performed before UpdateNode() finished. Is that correct?
If the above is true, we can eliminate this kind of racing by using topology to constraint disk type.
There was a problem hiding this comment.
No, there are other scenarios. We'll leave it as it is for now.
It's okay. We'll go over your changes when you file your pr. |
What type of PR is this?
/kind feature
What this PR does / why we need it:
Support reschedule disk with resource exhausted error in delay provisioning
Which issue(s) this PR fixes:
None
Special notes for your reviewer:
None
Does this PR introduce a user-facing change?
Additional documentation e.g., KEPs (Kubernetes Enhancement Proposals), usage docs, etc.: