My Unix team uses PowerHA with active-passive shared disk database.
as a dba I do the following steps to leave them the system ready in AIX.
1. install db2 and create the instance and the database on the active node.
2. Install db2 on the passive node. Copy the db2 port info from the /etc/services on the active node. They have to match on both sides, along with the instance name. Note that since your passive node does not see the shared file system until it is failed over, you wont be able to create the instance. However, the information in the /etc/service file will do it once powerHA fails over.
Power HA uses 2 script during the failover. One is to bring down db2 and clean up the processes (ipclean) before dismounting the resources and another to start db2 on the other server once the resources are mounted.
My unix team manages the switches manually using smitty. Unfortunately I don't have access to that unix part.
hope this help.
I'm not sure if I've understood you correctly.
There is a description of namely active / passive shared disk cluster using PowerHA in the Redbook I mentioned above.
Which type of cluster do you really need?
If you need active / active then what for?
ISAS/PDOA and pureScale both use TSA, not powerHA...
thanks for reply. In fact what I want to setup in my environment here is active/passive HA. The Primary and target servers each has own SAN Storage. Any issues happen to primary server including SAN Storage will failover to target server. From this setup architecture, should I still follow the same steps ? if you look at the redbook chapter 3, there's step to importvg, do I need to perform this ?tq
In a PowerHA active/passive cluster, how exactly is the DB2 global registry file (/var/db2/global.reg) (GREG) updated on the passive node to include the DB2 instance record from the active node?
According to the terse instructions in the above mentioned redbook (page 85, under "User/group setup and DB2 installation"), the instance is first created on the active node, then failed over to the passive node, where the instance directory (sqllib) is first deleted (?!?) and then the instance recreated (which "allows DB2 to automatically perform steps such as adding service
entries and modifying the db2nodes.cfg file for you").
Would it not be easier to do something (on the passive node) along the lines of:
/opt/ibm/db2/v11.1/bin/db2greg -addinstrec service=DB2,instancename=db2inst1,maintenance=1
Yes, that was a really strange recommendation - to install db2 on a standby node by removing the ~/sqlllib directory and creating an instance again on node2.
It's really much easier just to do the following steps on node2:
- install the db2 code
- create all db2 users and their groups with the same names, uids/gids, home
- run the command you mentioned above