NFS backups lock up VMs
- Thread starter externo6
- Start date
I'm using NFS for backup and never noticed such a problem, not saying it doesn't happen for me but I never noticed it as I have backups set to run when most are a sleep.
What are your server specifications, CPU, HDD, etc.
What compression are you using?
What mode?
What are your server specifications, CPU, HDD, etc.
What compression are you using?
What mode?
Intel Xeon E5-1650
64GB Ram
x2 Soft Raid Intel SSD DC S3500 (Most VM's are on)
x2 TOSHIBA DT01ACA200 2TB (Backups)
LZO compression with snapshot
6, but my understanding is that Prox goes though them one at a time though, may just stick to local storage and rsync the other backups across
Right, but back-to-back compressions can raise the load. I can't tell from the stats what the issue could be. Have you tried changing the compression/mode?6, but my understanding is that Prox goes though them one at a time though, may just stick to local storage and rsync the other backups across
Last edited:
I don't think you understand what LVM is. Using KVM for guests doesn't mean your host node shouldn't be using LVM.
http://pve.proxmox.com/wiki/VZDump
http://tldp.org/HOWTO/LVM-HOWTO/whatisvolman.html
If you aren't using LVM on your host node, and trying to backup guests via snapshot, it isn't going to work and you will run into issues such as what you are experiencing now.
Last edited:
In Proxmox 2.3 they implemented a new backup method for KVM which allows snapshots without using LVM.
Quote from the Introduction section:
Starting with V2.3, a new KVM live backup technology is introduced to Proxmox VE. The old backup system for KVM guests was using LVM snapshots which was working great in most situations. But it was not working if you store VM images on NFS, iSCSI LUN, Ceph RBD or Sheepdog. The new KVM live backup is targeted to work for all storage types and eliminates the need for special LVM configurations with enough snapshot space by avoiding the use of temporary storage. It is also very fast and effective because the new backup format is optimized for storing VM backups (sparse files, out of order data, minimized IO).
http://pve.proxmox.com/wiki/Backup_and_Restore
Also this is only regarding snapshots and NFS storage. Backing up to a local directory works fine.
Last edited:
ack;
as your backup works fine locally, it seems not to be a disk io issue;
you wrote the servers becoming unresponsive - does this only affect the vm's or the host system too?
how does your network setup look like?
is the nfs server on the same subnet?
do you use vlan's?
as your backup works fine locally, it seems not to be a disk io issue;
you wrote the servers becoming unresponsive - does this only affect the vm's or the host system too?
how does your network setup look like?
is the nfs server on the same subnet?
do you use vlan's?
The VM locks up (only the one been backed up) and the back up crashes, I have to kill the VZDump process. The NFS server is hosted in the same datacenter as the host, I they are both on the same subnet. They are liked via 1gig link.
Not sure if this is the correct way to really have offsite backups, if you dont think this is a good idea please shout at me