Add documentation for image-volume cache - Victoria
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Cinder |
New
|
Wishlist
|
Unassigned |
Bug Description
Hello,
We are running Openstack Victoria with external Ceph, and we have some use-cases where we deploy 100 volumes from images and if we don't use image caching it's taking a very long time.
After we implemented image-volume cache, the process it's working pretty quick but we have another issue. The cache volumes are eating a lot of space on the storage as these volumes are created almost at every deployment.
From the "image-volume cache" documentation https:/
"
Managing cached Image-Volumes¶
In normal usage there should be no need for manual intervention with the cache. The entries and their backing Image-Volumes are managed automatically.
If needed, you can delete these volumes manually to clear the cache. By using the standard volume deletion APIs, the Block Storage service will clean up correctly.
"
Is there a setting that will make the cache volumes to be deleted automatically? Because till this moment we always needed to delete them manually.
Do you happen to have more documentation about this topic?
Or explain us how it's exactly working. Because from what we see it's creating a volume cache for each cinder-volume service. We have cinder-volume service running on all 3 controller-nodes. And we saw that it's creating a volume cache for each of the controllers.
We are trying to automate this process, to see if we can make it to automatically cleanup by itself after each deployment, but we are trying to do this directly from Openstack and not with additional cron-jobs.
Let me know if you need more information.
Thanks!
Andrei Panescu
Changed in cinder: | |
importance: | Undecided → Low |
tags: | added: cache documentation image-volume |
Changed in cinder: | |
importance: | Low → Wishlist |
tags: | added: victoria |
summary: |
- documentation for image-volume cache + Add documentation for image-volume cache - Victoria |
Hi Andrei, I think this bug/question would be better aimed at the cinder team.