[Linux-cluster] updating ccs and running lock_gulmd

[Adam Manthei]

On Tue, Aug 16, 2005 at 03:49:20PM -0500, John Griffin-Wiesner wrote:
> I want to change the heartbeat allowed_misses setting.  I
> extracted the current config with "ccs_tool extract", modified
> cluster.ccs, then ran
> "ccs_tool create -O <dir of files> /dev/pool/cca"
> 
> Does lock_gulmd automagically see changes, or do I have to tell
> it to look?  "service lock_gulmd" offers only these options:
>   start|stop|restart|status|forcestop
> (I was hoping for a "reload".)
> 
> This is on a production system so I'd prefer to do this without
> causing everyone to be fenced off.
> 
> Thanks for any suggestions.

You can not change the heartbeat_rate or allowed_misses for lock_gulmd 
while the cluster is active.  In order to make the changes take effect, 
you will have to:

1. unmount all GFS in the cluster
2. stop all lock_gulmd in the cluster
3. modify ccs
4. start lock_gulmd
5. mount gfs

Note.  there is a bug that requires you to have to specify the
heartbeat_rate as a floating point number.  (see
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=166009)

Changing these values while the cluster is up will prevent new instances of
lock_gulmd (for both the clients and servers) from connecting to the active 
instance of the cluster, but it should not cause the running nodes to be
fenced.

-- 
Adam Manthei  <amanthei at redhat.com>