TOPIC:

Hi Salvatore,

I 'd like to pick your brain again. Sorry for the inconvenience

I'm looking at using DRBD to replicate my server to a separate geographic location as my data centre. The main site is a HA-Lizard deployment and I'm thinking of introducing new DRBD resources to sync off-site at block level because I don't want to maintain replication on each of the separate platforms that our hosting provides.

Have you any experience with this and/or can it be done within the scope of HA-Lizard deployment natively? It means adding DRBD resources but I don't want to break anything HA-Lizard because it is working like charm and I don't want to break it.

Not looking for any technical specifics, just some ideas/advice if you have any.

Cheers,

Nathan

Hi Nathan,
I have not tried this, but it should work in theory.

I have seen DRBD documentation that discusses "stacking" of resources to accomplish what you are looking to achieve. There are a few caveats however,
- DRBD over high latency links can be a problem. Linbit has a licensed version of DRBD specifically geared for deployments over a WAN.
- If you give it a try - do not add the new DRBD resource to the iscsi-ha config file (this will prevent iscsi-ha from trying to manage it)
- you will need to add filters to /etc/lvm/lvm.conf and /etc/lvm/lmaster/lvm.conf to prevent dom0 from reading the LVM signatures on the new DRBD device.

Perhaps it is something you can try with some test gear so as to not destabilize your production environment.

Also, my personal take on this is that it would be a good tool for faster disaster recovery, but, I would not rely on DRBD as a replacement for offsite backups. My view is backup should be done 3 ways, giving the maximum number of options should recovery be necessary.

- DRBD for near instant recovery with its block level replication
- regular data backups (files, DBs, etc...)
- VM image exports

Thank Salvatore,

That is exactly the input i was looking for... very helpful. Our company deploys backup solutions so we're real sticklers for redundancy and disaster recovery. This is solely a viability study at this point and is certainly not something we would implement without rigorous testing. I have been looking into Linbits WAN solution and geo clustering... block level replication really appeals to me and I'm hoping we can make good use of it as we expand our nodes... We will have decent fibre links to the main nodes so low latency will be mitigated as much as possible.

Cheers,
Nathan