Running DB2 9.1 on Linux . With HADR set up across 2 servers , the disk is not shared. I have a bunch of scripts which I execute , to force takeovers.

I want to now automate the monitoring and forced takeover of the DB2 servers. I know there are a few solutions out there including TSA, HACMP, Heartbeat for Linux . In terms of straightforward method , what would your recommendations be as to which method to use?

I would be interested in hearing your feedback

TSA is bundled with HADR license (AFAIK) and therefore that is the straight forward method. If you use any other monitoring tools in your environment, you can most certainly do an automatic takeover .. For eg. VCS provides a (sample) script for using HADR takeover .

HTH

Sathyaram

I'm running V9.5 on SLES11 and running HADR on 2 separate boxes with automated failover using a HADR script I modified from ftp://ftp.software.ibm.com/software/...rs/hadr_ha.pdf and using MONIT to monitor the heartbeat

Let me know if this is what you're looking for and I'll expand more

Thank you both for the replies.
bulump - are you able to expand on your initial explanation . 1)Do you have a 3rd server monitoring , or do you do all this on the 2 servers 2)How do you avoid the split brain scenario ? , for example , if server 1 becomes unavailable , and there is a takeover on server 2 - how do you avoid the situation where server 1 becomes available again, for example , if it hadn't been shutdown and it was a network failure. ?

I only have a 2 server setup....they monitor themselves via heartbeat on the hardware, network, and DB2 instance levels. If MONIT fails, it initiates the HADR sequence, pushing our Virtual IP and Primary DB2 responsibilities to the standby.

I handle STONITH by ssh-ing in the beginning of my HADR script to see if the other is Primary, if it is, the server falls into Standby. In the scenario where there is a catastrophic failure at the DC, I always have our DB1 server default to Primary...

Hi bulump!
You Install of db2 v9.5 on sles11 ok? I run it but mine appears to run into a GPF and java dump
for libdb2ure when creating the instance.
Tried several times (somehow a GUI gives the feeling it might work
next time) but it now fails consistently.

Hi bulump - how do you deal with the situation - of a)e.g network outage on SERVER 1 - and force a takeover on SERVER 2 standby 2)The SERVER 1 becomes available again and the DB2 on SERVER 1 still thinks it's PRIMARY - then you have a potential split brain scenario