[Linux-cluster] Cluster server maintenance

[Lon Hohberger]

On Thu, 2008-07-03 at 06:42 -0600, Gary Romo wrote:
> I have a two node cluster, RHEL 5, Protocol version: 5.0.1. Can anyone
> suggest
> the best method, and or explain -u and -q fo the clusvcadm command to
> me? Thanks!
> 
> Here is what I want to do:
> 
> 1. Shutdown the services running; DBs, apps whatever...
> 2. I don't want the services starting on the other node, or anywhere.
> 3. We don't want any fencing to take place
> 4. We do our maintenance; Patch server, whatever...
> 5. Bring the services back up; DBs, apps whatever...


> Now the only way I have found to do this so far is to disable the
> service.
> 
> # clusvcadm -d <service_name> (and maybe that is the only answer)

That's what it's for.

Stopping a service (clusvcadm -s) will stop the service until the next
member transition.

Disabling (-d) a service stops it until either quorum is broken or all
instances of rgmanager have been stopped.  That is, as long as one
instance of rgmanager is operating and the cluster is quorate, the
service will remain disabled.

Disabling autostart in Conga (or setting it to 0 in cluster.conf) for a
given service) means "on startup, treat this service as disabled instead
of stopped".


Locking rgmanager prevents failover, and is useful in mass simultaneous
shutdown operations, but less so for individual services.  The manual
page needs updating; '-l' only needs to be done once.

  [one node ] clusvcadm -l
  [all nodes] service rgmanager stop

-q = "don't print stuff"

-- Lon