[PVE-User] Ridiculous Live snapshot times with ceph
lindsay.mathieson at gmail.com
Tue Dec 2 05:37:42 CET 2014
Bit more detail - I examined the images in the pool while the snapshot
was going one.
- ID 301
- 1GB - 4GB RAM
- 1.3 GB actually used
- Snapshot took 8 min
- It created a image "vm-301-disk2" which was initial 8692M in size
- it was slowly filled with data
- After the snapshot was complete it was resized to 1365.99MB, which I
presume is the used memory map.
On 2 December 2014 at 12:34, Lindsay Mathieson
<lindsay.mathieson at gmail.com> wrote:
> nb. Starting the snapshot restore killed the running vm
> On 2 December 2014 at 12:32, Lindsay Mathieson
> <lindsay.mathieson at gmail.com> wrote:
>> Taking/Restoring live snapshots (test windows VM) is taking a
>> ridiculous amount of time - unusably so.
>> With images hosted on our NAS (NFS) or on gluster, snapshots take
>> around 30 seconds to take and/or restore.
>> With ceph:
>> - 9 *minutes* to take
>> - The restore has take *20 minutes* to get to 34%
>> And it generating iowait's of around 15%
>> Same hardware and VM's as previous tests, just different underlying storage.
>> Snapshots of a shutdown VM are near instant.
More information about the pve-user