TOPIC:

Hello!

Is it possible to have 2 iscsi-ha storages with own LV, DRBD and tgtd? I want both servers to run VMs as long as both are running. After failure, one of the tgtd would fail-over to remaining server with its VMs.

Thanks!
Lefty

In our standard setup (documented in the how-to document) you are able to run VMs on both hosts and live migrate with no restrictions. In the event that one of the physical servers fails, only the affected VMs will then be started on the remaining host.

So, if I am understanding your question correctly, there is no need to run 2 iscsi targets, etc..

You can, however, configure multiuple DRBD devices - each exposed as an individual LUN via TGTD. In this case, you should not have to run dual TGTD - just configure as many DRBD devices as needed making sure to configure them in iscsi-ha.conf and targets.conf.

Pulse Supply wrote: So, if I am understanding your question correctly, there is no need to run 2 iscsi targets, etc..

I want such setup because of performance. The IO of the secondary server has to be worse, when it has to pass all IO requests through primary server's tgtd.

Pulse Supply wrote: You can, however, configure multiuple DRBD devices - each exposed as an individual LUN via TGTD. In this case, you should not have to run dual TGTD - just configure as many DRBD devices as needed making sure to configure them in iscsi-ha.conf and targets.conf.

But I want two tgtds so that they can run on both servers and those servers access their discs through local tgtd.

I understand what you are trying to do. The SW is not designed to manage targets in that way on both hosts. You can consider increasing the number of links in the replication bond to get better performance from the slave.