DRBD9 - production deployment?

[pietrek]

Hi.

Would you guys consider DRBD9 with PVE 4.1 stable enough for a production ready solution with a two-node (no HA for now) setup hosting an email apliance storing few hundred GBs of data?
Or would it be better to stick to 3.4 until DRBD9 matures?

And if I go with 3.4 will it be enough deploy such a two node cluster? I don't see why it wouldn't be but I want to make sure.

Thanks!

 

[vkhera]

As far as I have been able to determine, DRBD9 will not work with fewer than three nodes. The third node does not necessarily need any data on it, but must be part of the cluster to provide quorum. I'm in this same situation as well, and am likely going to buy a tiny box machine for this purpose and then set up DRBD9.

 

[pietrek]

I don't think that third node is required for DRBD9 itself. From what I know, it's purpose is to make HA work and provide quorum for the failed-node-detection mechanism to function properly.
What I'm more intrested in is DRDB9's production environment readiness, general stability and performance. I'll be using LSI HW RAID with BBU and dual 10Gbit RJ45 interconnections between the nodes (if that's much of a factor).

I found out that VMware can do VT-X passthrough, so I'm going to test such setup on my laptop, but still - any feedback'd be appreciated
Thanks!