[Linux-cluster] Highly available services

[Jeff Sturm]

At a quick glance, I'd say fencing doesn't meet your needs, and RH
Cluster Suite may be overkill.  You've worked out the details of
failover, you just need to automate it.

Have you tried heartbeat?  It is well suited for simple failover of
resources such as IP addresses.

http://www.linux-ha.org/

-Jeff

-----Original Message-----
From: linux-cluster-bounces at redhat.com
[mailto:linux-cluster-bounces at redhat.com] On Behalf Of Soren Hansen
Sent: Saturday, December 15, 2007 6:07 PM
To: linux-cluster at redhat.com
Subject: [Linux-cluster] Highly available services

I've been trying to work this out from manuals and FAQ's, but I find
myself at a bit of a loss.

Scenario: I have 2 nodes, A and B. I want to set up a website using
apache/php and mysql. When both nodes are functioning, node A should
respond to http requests, and node B as database backend for node A and
possibly for other clients, too.

In case of node failure, the service of the failing node should
obviously fail over to the remaining node.

Both of the nodes are in fact running MySQL, with bidirectional
replication set up. No shared storage involved.

Both of the nodes are also running apache/php. The php and html files
and such are pushed from elsewhere to the two nodes, so no shared
storage here either.

Let's say that we have the following addresses:

a.foo.com   = 10.0.0.1
b.foo.com   = 10.0.0.2
www.foo.com = 10.0.0.11
db.foo.com  = 10.0.0.12

Node A will own 10.0.0.1, node B 10.0.0.2, and the node currently
providing www would also have 10.0.0.11, and the node currently
providing mysql would have 10.0.0.12, and the latter two could of course
migrate in case of a failing node.

Now, due to the nature of the services provided, there is no notable
potential for damage if a node fails, and the other node detects this
and the ip corresponding to its migrates to the still functioning node,
so there's not really any point investing a lot of cash in proper
fencing hardware. I just want to have as little down time as possible.

So, the question is: How do I do this using redhat cluster suite? Even
using fence_manual I still have to somehow notice the failing node, log
on to the remaining node, and do the fence_ack_manual dance, correct?

--
Soren Hansen
Ubuntu Server Team
http://www.ubuntu.com/