Hi All,

I have configured db2 hadr successfully and is working as expected. What I am not understanding is what happens when Primary server/instance goes down unexpectedly.

The client connections will flow to stand-by server as I have configure "ALTERNATE server", however the stand-by server database will be in "ROLL-FORWARD PENDING STATE" and database will not be readable

Any thoughts will help.

DB2 HADR by itself does not automatically failover to the standby. The following command must be issued on the standby:

db2 takeover hadr on db-name

If the existing primary server is not reachable by the standby, then you might have to issue this command:

db2 takeover hadr on db-name by force

You need a cluster manager to detect that a server failure has occured and that automatically issues the above commands. DB2 comes with TSA (Tivoli System Automation) that is included for free that will do this, but there are other cluster managers that can be used, or you can do it manually.

Thanks Marcus,

so does that mean I have to monitor disks, database(s) and the db2 services, and IP as failure of anyone of these can cause the non-availability.

or ping to a database is good enough to determine the service non-availability ?

You cannot ping a database, you can ping a server. You can connect to a database.

There are many strategies used to determine when to failover and issue the HADR takeover command, and some people just do it manually. The included TSA software will perform this function, but is a a little complex.

My company uses Red Hat Cluster manager to detect server problems, along with some scripts to detect when a database is not available. We also use a virtual IP instead of automatic client re-route, which is a little safer in ensuring that a split-brain situation never occurs.

There is a DB2 High Availability manual you get in PDF format that discusses these issues.