Multipath storage in cluster

[adonm]

Ok so I have a slightly complex setup as follows (IP's are simplified from actual):

server1 - master - 4x1gbit 10.0.0.1/24
server2 - slave1 - 4x1gbit 10.0.0.2/24

server3 - slave2 - 4x1gbit 10.0.0.3/24, 1x10gbit 172.16.0.3/24
server4 - slave3 - 4x1gbit 10.0.0.4/24, 1x10gbit 172.16.0.4/24

san1 - storagebox - 4x1gbit 10.0.0.5/24, 2x10gbit 172.16.0.5/24

And I'm trying to connect to nfs storage on the san in the cluster in such a way that server1&2 go through the 10.0.0.0/24 network and server3&4 connect to it via the 172.16.0.0/24 network

Ack so its doing my head in but I have dns setup so 'storage' resolves to 10.0.0.5 on server1&2 and 172.16.0.5 on server3&4 but after adding it on server1 I see that in cluster.cfg it adds it as 10.0.0.5, so server3&4 still go through the gigabit network, not the 10gbit fibre

Is there anyway in a cluster I can specify two different ways to get to the same storage or am I barking up the wrong tree? I guess if I can't I have to split the cluster into two, one for the 1gbit switched and one for the 10gbit crossover :/

 

[adonm]

Ah so I think I have a nice workaround, just putting my nfs mounts in fstab and testing now to see if it works (plan is to use shared directory storage rather than nfs, and mount the directories by specifying them on each server in fstab as an nfs mount)

seems to be working sofar, will update thread with status at end of tests

 

[dietmar]

Is there anyway in a cluster I can specify two different ways to get to the same storage or am I barking up the wrong tree?

No, that does not work. But you can use fstab as you do already.

 

[adonm]

Thanks dietmar,

Just wanted to confirm using fstab works at boot and for live migrations for NFS mounts, and it appears to mount exactly the same way (though still getting hangs when migrating vm's using virtio, that I can live with, considering my virtio vms boot in under 10secs though )

It would be nice to have a shutdown/restart migrate options for vm's with working acpi (maybe timeout the shutdown after 5 mins and just error out), rather than the 4 or so clicks it takes now but its not a priority ;P

 

[dietmar]

It would be nice to have a shutdown/restart migrate options for vm's with working acpi

But how do I test if a VM has working acpi?