I have a cluster of machines running RHEL 6.3 with the High Reliability Add-on.

I have configured a Clustered Logical Volume Manager (CLVM) file system for DB2, which is working correctly. Our DBA's have installed DB2 v9.7 on the cluster, providing me with a basic stop and start script. However, after a month of research, I have not been able to figure out how to set up the service within the cluster.

Can anyone point me in the right direction?

  • Are you trying to set this up with HADR via DB2? or something different? – Chris Aldrich Aug 7 '12 at 20:59
  • I am not familiar with DB2, or software associated with it, which is part of my problem. I don't know what HADR is. In working with a DBA, he was trying to use db2haicu (ibm.com/developerworks/data/library/long/dm-0909hasharedstorage) but it will not work for RHEL6. I am not certain if there is something we are missing, or if this is the way it HAS to be done and I just need to reinstall and older version of RH and resetup the cluster. – Scott H Aug 7 '12 at 22:05
  • You only need db2haicu if you want automatic failover. Otherwise you can implement regular clustering/failover using just simple HADR in DB2. However, db2haicu is a LOT trickier to set up. At our company we are going to start with just plain HADR and put in db2haicu at a later time. Do you know what your needs are? What is required of you? – Chris Aldrich Aug 8 '12 at 12:36
  • The goal is to have a data warehouse cluster with 2 nodes and failover. If the node connected to the DB fails, then the other node mounts the file system and takes over. Just a simple 2 node failover cluster. As I said, having no db2 experience, I am sort of at the mercy of our DBA's to provide direction, and we both have our idea's of how we think it should be done. – Scott H Aug 8 '12 at 13:16

As I mention in my comments above, setting up HADR (High Availability Disaster Recovery) with DB2 is pretty simple, especially if you are going Active/Passive with your nodes. You do not gain automatic failover per se. You would use the db2 takeover hadr on database <database-name> command for a normal takeover for maintenance and a db2 takeover hadr on database <database-name> by force in the case of a hard fail of the active node. This article is a good one for how to set that up. (I know it is about Rational Jazz, but it has the steps for how to configure simple HADR without db2haicu).

It sounds like you want automated failover. This means moving up the levels of HADR to start using db2haicu and TSA. Ember Crooks' blog has a part 1, part 2, and a part 3 that shows to set this up. This gets a little trickier because it involves another layer between the OS and the database that needs to be configured and it must know about IP addresses, ports, a virtual IP address, etc. But the benefit is that this extra layer checks to make sure your active nodes stay active, and if not, it fails over to a passive node.

I would also recommend (you and the DBA's) watching this replay from the DB2NightShow. Vikram Khatri from IBM goes through a set of slides that take HADR from its simplest to its most complex, based on what you are trying to do (ie, from active/passive on the same LPAR all the way up to multiple active/active phyiscal nodes with load balancing, etc.)

I'm guessing between Ember's blog and the DB2NightShow link you'll find what you are looking for. However, don't discount the first link. You could set up simple HADR without db2haicu and always "automate" failover with some scripts or something. It might be simpler than db2haicu until you can figure db2haicu out.


just put the simple script as a resource on your RHEL Cluster. Since your on CLVMD I assume that clvmd daemon are both active on your nodes. If node one goes down RHEL Cluster will mount all active VG and LV on the other nodes as per your FAIL Over Domain and then start the database using the script provided to you. THe script should be on both server on same location for this to work.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.