On 18/07/13 00:12, Martin Langhoff wrote:
On Wed, Jul 17, 2013 at 9:34 PM, Digimer <[email protected]> wrote:
The easiest, native way under RHEL/CentOS is to use corosync + cman +
rgmanager. The configuration you are describing will be simple and will be
properly supported until 2020 (at least), and not need hacks.
If you're interested in this approach, I can help. Here or on #linux-cluster
on freenode's IRC.
Thanks for the offer to help. Is there any clear setup guide you can
point me to?
My TZ is EDT, so midnight (bedtime!) now. I won't be awake and on
email/irc until tomorrow morning.
Heh, same timezone, but I'm more of a night owl. :)
I have a tutorial that was written for people who want to host
highly-available VMs on a two-node red hat cluster. It goes into a lot
of detail that you may not be interested in, but I think it's pretty
comprehensive (I tried to assume no prior knowledge of HA). So perhaps
you can tease out the parts you're interested in.
https://alteeve.ca/w/2-Node_Red_Hat_KVM_Cluster_Tutorial
You're configuration would need basically;
* Node definitions with fence methods defined
* Resource section covering your storage and daemon
* failover domain to control which node is primary for a given service
and which is the backup
The tutorial covers clustered LVM and uses the GFS2 clustered file
system. So it anticipates a somewhat complex setup. If you are looking
for simple failover, you can skip all of that. You could even dump LVM
all together, if your goal is to simply support MySQL's data storage.
So the config, in this case, you be;
* The cluster name is foo
* This is a two node cluster (disable quorum)
** Node 1 is this, and here is how you fence it
** Node 2 is this, and here is how you fence it
* Resource;
** I have a file system resource call X mounted at Y
** I have a script resource that controls daemon Z
* Failover Domain
** I have an ordered domain that says "run on node 1 when possible, node
2 otherwise. if you fail over to node 2, stay there when node 1 returns
* Service
** Create an ordered service that follows the rules set in failover
damain. This service requires the FS to mount before the daemon service
starts. Stop in the reverse order
That's it. It might seem a little overwhelming at first, but it really
is pretty simple. You already understand the concept of fencing, which
trips up most people, so you're more than half-way there. So long as
your switch handles multicast, your golden. If not, no big deal, just
add the configuration option that forces unicast mode.
hope this helps
--
Digimer
Papers and Projects: https://alteeve.ca/w/
What if the cure for cancer is trapped in the mind of a person without
access to education?
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems
