Backup of VMs with problems

[cesarpk]

Hi to all

To anyone that can help me with my Backups:

My scenery on the LAN:
For Backup:
- I have a Centos 6.4 x86_64 as backup Server through NFS, two NICs Gb/s with balance XOR
- All nodes (PVEs and Centos) have NICs of Gb/s
- All Backups of VMs are "Snapshot"

Hosts PVE with VMs:
- I have four PVE 2.3 nodes with some VMs, each node with bond active-backup (for the moment i can't upgrade to PVE 3.1)
- I have a single PVE 1.8 node with some VMs, the host with bond balance-alb (for the moment i can't upgrade to PVE 3.1)
- All the Virtual Disks of the VMs are local

Start time of the backup:
For all PVE nodes: 00:00 hours

My NFS configuration on all the nodes:
- In the NFS Server:
/backups/vm xxx.yyy.zzz.a/32(rw,sync,no_subtree_check,no_root_squash)

- In all PVE Nodes:
Using PVE GUI, then "mount" shows:
A) In all PVE 2.3 nodes:
xxx.yyy.zzz.a:/backups/vm on /mnt/pve/Server-Backup1 type nfs (rw,vers=3,addr=xxx.yyy.zzz.a)

b) In PVE 1.8 node:
xxx.yyy.zzz.a:/backups/vm/dump on /mnt/pve/Server-Backup1 type nfs (rw,addr=xxx.yyy.zzz.a)

The Problem:
- The PVE 2.3 nodes do backup without problems
- The PVE 1.8 node starts the backup, but after some time vzdump is frozen, after in this node i have that kill the vzdump process, unmount and remove the LV created by vzdump

Analysis extras:
- During the freezing of vzdump in PVE 1.8, if i try to connect by ssh to PVE 1.8, the connect is quick and successfully

- During the freezing of vzdump in PVE 1.8, if i since this node connected by ssh run "ls -l" to the NFS resource shared, the screen don't shows the files, and [ctrl - c] not is useful for come back to prompt and the screen shows "^C"

- During the freezing of vzdump in PVE 1.8, if i since a PVE 2.3 node connected by ssh run "ls -l" to the NFS resource shared, his screen shows the files after of some seconds

- After, in the morning, if i since the node PVE 1.8 connected by ssh run "ls -l" to the NFS resource shared, the screen shows quickly the files and the prompt come back

- Before i have the backup Server, "PVE 1.8" always did ​​backups successfully to another node PVE

The Question:
1- Where are the logs for revise this problem?
2- How do I prevent this freezing to achieve a successful backup?

I will be very grateful to anyone who can help

Best regards
Cesar

 

[SamTzu]

There has been long standing problem for vzdump (read RSYNC) backup's over NFS. (LVM2 might be part of the problem.)
No body seems to know what the problem is but most agree it's kernel related (and/or big file related.) There are tens of threads here about similar related problems on this forum alone.
These days to be safe you should backup to localhost then after - move the backup to the NFS share in question. That works better.

The day btrFS comes of age and Proxmox starts to use it I will do standing ovation. LVM snapshots "suck" major time and I personally see that as major problem with vzdump backup in Proxmox.

 

[cesarpk]

There has been long standing problem for vzdump (read RSYNC) backup's over NFS. (LVM2 might be part of the problem.)
No body seems to know what the problem is but most agree it's kernel related (and/or big file related.) There are tens of threads here about similar related problems on this forum alone.
These days to be safe you should backup to localhost then after - move the backup to the NFS share in question. That works better.

The day btrFS comes of age and Proxmox starts to use it I will do standing ovation. LVM snapshots "suck" major time and I personally see that as major problem with vzdump backup in Proxmox.

Thanks SamTzu for your answer.

Nowadays my backups are working perfectly.

For anyone who has practiced the same problem and needs fix it:

My problem is that i did use the parameter: --dumpdir <the directory of NFS mount>
instead of: --storage <name of the storage NFS>

Best regards
Cesar