Redundant GlusterFS servers as Storage - WHY NOT!?

[firegs]

Can anyone tell me why we cannot use more than a single IP for a GlusterFS storage device? Isn't that one of the central aspects of GlusterFS - that its multiple servers and not just one?

I have a two-node GlusterFS 3.4.1 cluster running. I can only add one of the two node's IPs to Proxmox. If I take that node down, the storage is down, yet node #2 was more than available to take care of business.

I've tried rrDNS, and while it works (sort of) with ping, every other ping fails when a node is down - and Proxmox still says it either cant mount, or the Storage is offline.

Will there ever be a native GUI fix for this? If not, what was the point of adding GlusterFS support?

Sam

 

[firegs]

Ugh. It was rrDNS. As soon as I went back to not using rrDNS, it works fine. Close thread.

 

[fknorn]

Hi there, sorry for digging this out - but how did you manage to get the redundancy working? If you can add but one IP through the proxmox GUI?

 

[firegs]

Hi there, sorry for digging this out - but how did you manage to get the redundancy working? If you can add but one IP through the proxmox GUI?

Say I have two GlusterFS nodes - 10.0.0.220 and 10.0.0.221 - If I simply add the 10.0.0.220 node in as a Gluster Storage type, the glusterfs api intelligently finds the other node. You dont need to tell Proxmox about both. Just one. Works great!

 

[fknorn]

Sweet, thanks, indeed, I just tried it out!