Can I do the GlusterFS setup with a LB or heartbeat service in SLES 12? If yes, How?

Question

I've installed GlusterFS on Suse 12 with 2 VMs & 2 replica. When one server goes down it doesn't failover to the 2nd IP automatically. Can I do the same with a loadbalancer or heartbeat service? If yes, How? I need HA configured for GFS.

Gluster volume info: Volume Name: gv0 Type: Replicate Volume ID: 3b2eb6c9-a869-4e72-8cec-7d04583ea1d8 Status: Started Number of Bricks: 1 x 2 = 2 Transport-type: tcp Bricks: Brick1: 10.116.33.146:/export/sdb1/brick Brick2: 10.116.33.134:/export/sdb1/brick

Gluster peer status: Number of Peers: 1 Hostname: mo-2005258b4.lab-rot.saas.sap.corp Uuid: 85f4907b-d142-4fea-acc0-83ee74116142 State: Peer in Cluster (Connected) Other names: 10.116.33.134

Answer 1

Heartbeat of a load balancer shouldn't be used or wanted in this case. Yes, you are mounting using a single IP or hostname from a client (via the mount command or via autofs), but that's ONLY during the mount process. After this, Gluster is handled via FUSE on the client side, and is aware of (and connected to) all storage nodes. You're certainly not failing over for a different reason.

Would you mind posting your gluster info so that we might check it out? A lack of failover in this case could be any number of other problems. MAKE SURE NTP IS WORKING. These nodes absolutely have to be in time sync in order to function properly as Gluster peers.

Here are some commands you can run to output some useful information:

gluster volume info all
gluster peer status

You can also find the log for your volume / brick via this:

gluster volume log locate <VOLNAME> [BRICK]