Volume retype does not have any visibility into its progress
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Cinder |
Incomplete
|
Undecided
|
Unassigned |
Bug Description
When performing a volume retype of a large volume, I have hit cases where the token expires and causes the volume to be stuck in migrating/retyping.
- Cinder logs on verbose and debug do not show anything relevant on this volume id (nor the volume id of the new temporary volume it creates)
- If the volume is attached to a vm, I can see logs in nova about authorization failures. However if the volume is detached that won't be the case.
- In ceph, the only visibility I seem to have is by checking for any active IO on the destination temporary volume.
It would be nice if volume status itself could possibly show some sort of progress on this retype or if it failed due to a token expiration. Same goes for cinder logs.
This above was done on a focal ussuri deployment using rev 308. (16.4.2)
Can you offer a bit more in the way of details, I'm assuming a Ceph backend, how are things configured and connected, how many volumes, what is the load on the deployment at the time of failure, volume specs, etc. Eric mentioned in our bug meeting https:/ /etherpad. opendev. org/p/cinder- bug-squad- meeting that it shouldn't be a problem if things are delployed correctly - how was this deployed? Etc... Could be a bug, but more information would be most helpful.