[Linux-HA] Starting a single instance of a clone after a failure?
beekhof at gmail.com
Thu Jul 19 04:14:42 MDT 2007
On 7/18/07, Jeremy N Thornhill <jeremy.thornhill at duke.edu> wrote:
> I am trying to use heartbeat to manage a multi-master MySQL configuration
> that currently consists of 2 nodes. Both nodes should be functionally
> equivalent, and MySQL should be kept running on all nodes as much as
> possible to keep changes synchronized.
> To do this, I have created a clone resource to keep MySQL running on
> multiple machines. I then added an IP resource which is colocated with
> the MySQL clone resource.
> This works as I would expect - when heartbeat starts, both of the MySQL
> nodes start. if a MySQL resource on a node is made to fail, the IP
> resource moves to the other node which has a running MySQL instance.
> After this has occurred, I would like to restart the failed MySQL server
> while leaving the newly active server untouched. However, as far as I can
> tell, there is no way via either the GUI or crm_resource to force a single
> instance of a clone back online after a failure. The only way I have
> found to restore the original node is to do a "resource cleanup," which
> does work, but causes *both* mysql clones to restart. This bounces the
> daemon on the active node, which interrupts service.
> So, I have a few questions:
> 1) Is it possible to manually restart a single instance of a clone
you tried crm_resource -C -r mysql:0 -H hostname ?
note the ":0" this is the way to address a specific clone instance
> 2) Is there a way to have the "cleanup" bring the failed node back but
> *not* restart the active node?
clone handling has been significantly improved in recent versions (and
even again since then), you might want to try grabbing the latest code
> 3) Is there a better way to configure heartbeat to deal with multi-master
master/slave resources can have as many masters as you like
More information about the Linux-HA