RE: Do you have much experience with "activeversion" not being at the right level?

MOS Doc ID 1176853.1 is with regard to failed upgrade from 10gR2 to 11gR2, but, does mention getting an error on 'crsctl set crs activeversion'. It references some files that need to be cleaned up from $GRID_HOME/cdaa/crs/ directory. (See the doc for details.)

Again, I know the versions are different, and it's not exactly what happened to you, but it may offer some clues.

Also, this is off the top of my head, and may actually end up being destructive, so, think twice before typing it, but, if the only problem is the activeversion on that node, is it possible for you to manually run 'crsctl set crs activeversion' on that node? Not sure if you are allowed to do that, or if it's a good idea, but it *seems* like it would do what you need it to do. If that fails, the files mentioned in the MOS Doc previously referenced may need to be cleaned up....

I know, not a very complete answer, but perhaps you'll find something useful there?

-Mark

-----Original Message-----

From: oracle-l-bounce_at_freelists.org [mailto:oracle-l-bounce_at_freelists.org] On Behalf Of Tyfanie Wineriter Sent: Monday, April 11, 2011 3:55 PM
To: 'oracle Freelists'
Subject: Do you have much experience with "activeversion" not being at the right level?

Good Afternoon all,

I have been struggling with a difficult situation for a couple months now, and I was wondering if anyone on this list has seen anything similar and what worked (or didn't work) for them.

In January, I upgraded our clusters Grid Infrastructure from 11.2.0.1 to 11.2.0.2. On one production cluster, the upgrade went perfectly. Unfortunately, on my last 4-node production cluster, ASM crashed in the middle of it, on all 4 nodes, and it would not come up with 2 separate versions. This required Oracle Support, who had me shut down the 2 upgraded 11202 nodes & do the last 2 11201 nodes alone. I did this, and checked the releaseversion after every one, and it looked successful. The cluster ran out of the new 11202 Grid_Home and that's how we've run since January.

That is, it *looked* successful, until I tried to install the 11202 Database client last month. I then discovered that Node3, though it updated it's releaseversion, did not complete it's upgrade fully, and thus did not inform the OCR of what it was at. My activeversion was never set to 11202, which is stopping me from installing the 11202 database client.

The grid has run great otherwise, out of the 11202 New_Grid_home, and the months of doing what Oracle Support suggests (including trying to downgrade Node3 back to 11201) has not worked, and we are hesitant to make grid-wide changes (like restoring the OCR to the 3-month-old OCR) when the cluster is working well otherwise.

Now for my actual question:
  We have decided to remove the trouble node out of the cluster & then add it back in.
  My concern is that just removing it & adding it back in will not upgrade the activeversion.
  Has anyone done anything like this, when their nodes were out of synch? What were your results?

Thanks in Advance,

~ Tyfanie Wineriter ~

Database Administrator
University of Oregon

--

http://www.freelists.org/webpage/oracle-l

--

http://www.freelists.org/webpage/oracle-l Received on Mon Apr 11 2011 - 15:14:54 CDT