AnsweredAssumed Answered

DB2 with HADR -- issues with DIVDB table

Question asked by daryl.brown_ACI on Feb 14, 2018
We're using a DB2 database with HADR replication.
One issue we've run across is that when we go to failover to our second node -- which happens nice and smoothly; it's a roughly 5-second blip in database connectivty -- we run into problems with the DIVDB table.
In particular, when you run certain utilities (e.g., ucybdbar/ucybdbre/ucybdbun), the following command is issued:

“ALTER TABLE DIVDB ACTIVATE NOT LOGGED INITIALLY WITH EMPTY TABLE”.

In HADR, that causes problems when you failover to the backup node, as you then get errors like this:

SQL1477N  For table "USRUC4.DIVDB" an object "4" in table space "8" cannot be accessed.  SQLSTATE=55019

(This is documented behavior by IBM; DB2 is working as designed in these cases.)

In terms of the AWA itself, at worst, this error can prevent you from being able to start up CP/WPs, and at best, it just causes problems running the same utilities that access this table.

I've been talking with Automic Support about this, but their solution is to simply drop and recreate the DIVDB table (which is basically what IBM advises for dealing with this issue).  While that's doable, that's a manual step that now has to be executed as part of a failover, and is therefore less than desireable.

Has anyone else encountered this issue, and/or identified a good solution to avoid having to manually delete and recreate this DIVDB table each time?

Thanks in advance!

Outcomes