linux-ha at ragnark.vestdata.no
Wed Oct 16 12:59:51 MDT 2002
On Wed, Oct 16, 2002 at 08:32:55PM +0200, Daniel Khan wrote:
> > > Unfortunately the whole thing produces much CPU load which
> > slows down the
> > > server and my question is
> > > if it would help if I used rsh and/or run rsync in daemon mode.
> > Most likely. (If it is the ssh-process that actually is producing the
> > CPU load)
> Well it's rsync with about 35% and ssh with about 5%.
> It's a SMP machine - so it's not such a catastrophe.
> I'll try the without encryption thing - thanks!
Then changing ssh has a potential of only 5% improvement.
My guess is that your server is IO-bound. You need faster disks / RAIDs.
> > > Is there a way to add a resource "gracefully"?
> > With nice_fallback on:
> > 1. Change the configuration on your standby-server
> > 2. Restart heartbeat on the standby-server
> > 3. Shut down heartbeat on the primary-server, the services will move
> > to the standby-server.
> > 4. Change the configuration on the primary-server
> > 5. Start heartbeat on the primary server.
> Well the problem it's an active-active setup.
> Node 1 is webserver and failover database server
> Node 2 is database server and failover webserver.
Sorry, I didn't see that.
> Wouldn't it be nice to have a command like the reload command
> that checks haressources and takes all actions to get all ressources
> listet there?
Yes, it would be nice.
It's not _that_ easy to do though - you would need to examine the old
configuration as well as the new one to figgure out what resources
should be stopped and what should be started. No rocket science exactly,
but it takes a little bit of coding.
More information about the Linux-HA