Bug #3521
one should use rbd snap to create a consistent image
Status: | Pending | Start date: | 01/22/2015 | |
---|---|---|---|---|
Priority: | Normal | Due date: | ||
Assignee: | - | % Done: | 0% | |
Category: | Drivers - Storage | |||
Target version: | - | |||
Resolution: | Pull request: | https://github.com/OpenNebula/one/pull/45 | ||
Affected Versions: | Development |
Description
Hello,
already suggested this at https://www.mail-archive.com/users@lists.opennebula.org/msg16746.html
I tested if "rbd copy" is creating a consistent (in time) copy of the source image.
What I did:
- created a 20G rbd-vol
- filled rbd-vol with random data
- generated md5sum of rbd-vol
- started copy of rbd-vol to rbd-copy
- while running, started to overwrite the source-image with random data
- after copy finished, generated md5sum of rbd-copy
==> md5sum's differ
If I use a snapshot and copy the snapshot to the target-vol, the md5sums match.
I will create a merge-request
History
#1 Updated by Fabian Zimmermann over 6 years ago
#2 Updated by Ruben S. Montero over 6 years ago
Hi Fabian,
Thanks for this. On a second thought, OpenNebula does prevent this (i.e. an image in use (USED state) is not removed and so on...) So, did you have any corruption problems in the past? Maybe we have an issue at oned level that needs to be solve...
Cheers
#3 Updated by Ruben S. Montero over 6 years ago
- Pull request set to https://github.com/OpenNebula/one/pull/45
#4 Updated by Fabian Zimmermann over 6 years ago
Hi,
Ruben S. Montero wrote:
Thanks for this. On a second thought, OpenNebula does prevent this (i.e. an image in use (USED state) is not removed and so on...) So, did you have any corruption problems in the past? Maybe we have an issue at oned level that needs to be solve...
I don't think there is a problem in oned, it should work like expected. I just worry about this single/special event if you create a "hot" hdd-snapshot while the vm is running.
I know the manual tells "you are responsible to avoid data corruption during snapshot-creation", but if ceph provides a clean (and simple!) way to reduce the risk, we should use it, isn't it?
Fabian
#5 Updated by Ruben S. Montero over 6 years ago
- Category set to Drivers - Storage