Two node proxmox cluster with ZFS and DRDB support

[unni]

Hi,

I have created a two node proxmox cluster and configured ZFS, DRDB

I followed this documentation.

http://jsmcomputers.biz/wp/?p=559

I noticed, there is an option to add ZFS directly on the web GUI. Do that method supports a live migration? Do I need to mixup drdb and ZFS to get enable live migration?

 

[unni]

Somebody please advice

 

[liska_]

I am not 100% sure about live migration, but I think that this zfs option uses iscsi and in your setup there is used lvm.
I had some problems with drbd previously and creating manually every drive is a no-go for me, so I would choose glusterfs on zfs direction for example. But it depends on your demands, as it could be a little slower.

 

[wolfgang]

No the ZFS on the Gui is not for ZOL it is for ZFS over iSCSI like FreeNas, Nexenta ....
So it is not possible to use it this way.

 

[unni]

I am not 100% sure about live migration, but I think that this zfs option uses iscsi and in your setup there is used lvm.
I had some problems with drbd previously and creating manually every drive is a no-go for me, so I would choose glusterfs on zfs direction for example. But it depends on your demands, as it could be a little slower.

Thanks Liska for the reply.

Yes. In this setup. LVM is on the top of ZFS. Will Glusterfs on ZFS gives the same functionality of the setup? If yes, I can test out those. I am trying to find a stable way to build a cloud in proxmox.

 

[liska_]

Well, it depends on what you are looking for. It is a distributed filesystem, it has file snapshot as well as volume snapshot. I tested it and it seems to me trustable. When there is a split brain, you can solve it easily on file level and not volume level, which seems to me really great as it locked down just problematic vm files.

 

[unni]

Is there any way to use ZFS without DRBD for live migration? I have read about something named as zreplicate.

 

[wolfgang]

I only know zfs send/receive, but this is not for HA it work like an asynchron replica.
so when your Host failed you must turn on the Vm manually.
But you need no shared storage.

 

[liska_]

As Wolfgang has written you will need some non-free products like Nexenta etc for sync HA. There were some instruments for setting HA ZFS by Sun, but I found them very hard to deploy and use. Zreplicate use send and receive, this means you can do a snapshot every five minutes for example and achieve nearly realtime backup.
But this is the part where glusterfs come to solve distribution of data between nodes very easily in comparison to drbd.