Greetings,

I'm trying to setup an active-active two-node cluster with shared
storage for mail (Postfix and Dovecot) and I would like to know which
are your recommended procedures regarding it. The objective of this
cluster is having both load balancing and high availability.

I have a SAN and two nodes connected to it through Fibre Channel. I'm
able to format volumes in the SAN from any of the nodes and I can mount
the filesystem in both machines. But then, of course, if one of the
nodes write to the filesystem, the other would not be aware.

Besides using file servers such as CODA and NFS, I was considering to
use a distributed filesystem such as OCFS2 or GFS. I actually did it,
but then I run into this problem: when I unplug the network cable in one
of the nodes, the other one also hangs, so then I end with a
zero-availability solution :p

I'm aware of the fencing processes both OCFS2 and GFS use, and I'm using
the two-node mode of GFS, but I still can't make what I want: having two
nodes writing to a shared storage, and keep one node writing to the
shared storage if the other fails.

Of course, I have a load balancing solution with lmonitor running on top
of the two nodes, which won't forward any more work to the faulty
machine. The load balancer is also highly available, since there's a
heartbeat running and it has a companion. So the whole cluster looks
like this:

| lb1 -- lb2 (active-passive)
|  switch
| node1 -- node2 (active-active)
| SAN

I've read documents with study cases regarding A-A two-node
configurations, and they seem to use the same tools I'm using, but I
don't get it yet. Is there anybody around with working recommendations
regarding this arrangement?

Thank you very much in advance,
Jose
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems

Reply via email to