Bringing down and up a HA cluster

Hello Is there a preferable/right order to bring down and up a HA cluster (primary / HDR / RSS) ? Also is it better to pass through quiescent mode, forcing a checkpoint,... Thanks ------------------------------ Samuel ------------------------------ #Informix

Hi, when shutting down ANY kind of Informix server, a checkpoint makes sense in order to minimize the startup time later on. The startup routine will start a recovery process which begins to recover from the latest checkpoint. This might take significantly longer if the checkpoint is long ago. That way, you can minimize the downtime by initiating a checkpoint right before the shutdown. My preferred way to shut down an instance is onmode -ukcy, no matter what role it has. This does work also on a secondary instance, even though quiescent mode makes no real sense on this kind of server. Best, Marcus Haarmann

Are you using connection manager to manage the failover in your cluster ? If so, then shut down servers in reverse order as specified in your FOC definition So if FOC ORDER=PRI,SDS,RSS then shut down RSS, then SDS, then PRI Regards Rainer ------------------------------ Rainer von Bongartz ------------------------------

Thanks ------------------------------ Samuel To ------------------------------

I don't use a connection manager. So there is no preferable order to shutdown the servers PRI,HDR,RSS, just the down and up have to be in reverse order ? ------------------------------ Samuel To ------------------------------

Shmuel: No, there is not specific good or bad order to shutdown an HA cluster. Personally I would shutdown the primary first to give the secondaries a few moments to process the last couple of log buffers sent to them, but that's a minor thing. They would just process those logs on startup if they were saved or request them to be resent if they were not on disk yet at shutdown. There is no reason to do a separate forced checkpoint before shutting down. As long as we are talking about a proper shutdown using onmode -ky or -uky that always triggers a checkpoint as part of the shutdown. Art ------------------------------ Art S. Kagel, President and Principal Consultant ASK Database Management Corp. www.askdbmgt.com ------------------------------

If you use onlog you should see a checkpoint record as the last thing the engine does but I always use '- yuck' as it just easier to remember Cheers Paul Paul Watson Oninit LLC +1-913-387-7529 www.oninit.com Oninit® is a registered trademark of Oninit LLC

thank you all ------------------------------ Samuel To ------------------------------

Rainer I believe the ORDER sub-parameter defined in the Connection Manager cmsm.cfg file is now deprecated, and Connect ion Manager adopts the Primary node's $ONCONFIG HA_FOC_ORDERparameter setting, which still has the default value... [ informix@informix-01 ] ~ > onstat -g cfg | grep HA_FOC_ORDER HA_FOC_ORDER SDS,HDR,RSS ...although that doesn't make a blind bit of difference in the context of this thread. ------------------------------ Kirit Rana ------------------------------