Are live backups working on DRBD9 storage?

[dendi]

Hello,

I'm testing DRBD9, backup works but I noticed this line:

INFO: snapshots found (not included into backup)

I know that drbdmanage has not all functions working (see disk expansion), so are snapshot working?

I typed this command during backup:

drbdmanage list-snapshots
Snapshot list is empty

INFO: starting new backup job: vzdump 100 --node pve213 --remove 0 --compress lzo --storage nfsbk --mode snapshot
INFO: Starting Backup of VM 100 (qemu)
INFO: status = running
INFO: update VM 100: -lock backup
INFO: backup mode: snapshot
INFO: ionice priority: 7
INFO: snapshots found (not included into backup)
INFO: creating archive '/mnt/pve/nfsbk249/dump/vzdump-qemu-100-2016_05_07-11_45_50.vma.lzo'
INFO: started backup task '7bcf2207-4bb5-4450-b5bb-11b8c953be97'

 

[dietmar]

Please can you post your VM config /etc/pve/qemu-server/100.conf?

 

[dendi]

yes, thank you:

bootdisk: virtio0
cores: 2
ide2: nfsbk:iso/CentOS-7-x86_64-Minimal-1503-01.iso,media=cdrom
memory: 4096
name: demo
net0: bridge=vmbr0,virtio=32:64:37:38:34:35
numa: 0
ostype: l26
scsi0: drbd1:vm-100-disk-1,cache=writethrough,discard=on,size=80G
scsihw: virtio-scsi-pci
smbios1: uuid=c8795fea-c650-4e52-a659-f983b5cfc4c5
sockets: 1
vga: qxl

[PENDING]
boot: cdn
bootdisk: scsi0

 

[dietmar]

OK, you found a bug. I filed a bug report here:

https://bugzilla.proxmox.com/show_bug.cgi?id=987

The good thing is that you can ignore that warning. It was falsely injected because
you have pending changes. If you stop/start the VM it will be gone.

 

[dendi]

Happy to help with debugging

I stopped and restarted the VM and the "INFO: snapshots found (not included into backup)" disappeared during backup.

But during backup should'nt I see a snapshot with drbdmanage list-snapshots or lvdisplay?

 

[dietmar]

But during backup should'nt I see a snapshot with drbdmanage list-snapshots or lvdisplay?

No, because we use a special KVM feature to make snapshot backup. So there is no need to create storage snapshots.

see https://git.proxmox.com/?p=pve-qemu...5250ca9f681e5377bd1821479259e81d4e78d;hb=HEAD

 

[jasminj]

No, because we use a special KVM feature to make snapshot backup. So there is no need to create storage snapshots.

see https://git.proxmox.com/?p=pve-qemu...5250ca9f681e5377bd1821479259e81d4e78d;hb=HEAD

I played now with two variants of a DRBD9 setup.
First I created it the "old way" and defined a thin pool on top of the DRBD device. I added this as storage on the WebGUI and was able to create a VM and was also able to create snapshots.
BUT, it was not possible to define the storage as shared and I need this, because I want to make my two node cluster with it.
Then I read your post and you write "So there is no need to create storage snapshots". I deleted the VM, the storage and the thin LVM pool. Then I created a normal LVM group on top of the DRDB device and added this as storage. Now I could define it shared, but I can't make snapshots now!
Now I am confused, why this is not possible? When I read the link above, it should be technical possible to create snapshots on a DRDB9 backed up shared LVM storage, even LVM doesn't support snapshots in this configuration.

And yes, I am new to DRDB9 and HA clusters, but I read a lot in the meantime, like this, this and this. I have reasons to use it the "old way" and not using drbdmanage:

• Mainly because I want to avoid an additional LVM layer underneath DRDB9 by using partitions/disks for each DRDB9 resource.
• DRDB9 support is currently in beta state in Proxmox.
• I want to have dedicated DRDB9 resources for each of my HA machines to simplify Spilt-Brain recoveries (as mentioned above on different a disk/partition).
• DRDB9 should run directly on a disk, so that I can easily access the data if something gets really wrong and I don't need to deal with LVM on top of LVM.
  
• I have another DRDB9 resource (and disks) on my machines, which is not managed by Proxmox (a kind of shared NAS provided by the host).

BR,
Jasmin

 

[jasminj]

After playing for another 5 hours, I discovered that I don't need snaphots, but backups.
And this is working with my DRDB9 backuped LVN storage. This thread was very helpfully.
Now there is only one thing left. After a server restart, the DRDB9 daemons are not automatically started and my server is not working as expected. I created another thread for this issue already, but got no answer till now.
Maybe someone of the Proxmox team has time to answer my question in the other thread.
BR,
Jasmin