Patching cloud control from 12.1.0.1 to 12.1.0.3 - my experiences

Got round to doing this today and thought i'd share some of the details of what happened during the upgrade.

I followed the oracle doc http://docs.oracle.com/cd/E24628_01/upgrade.121/e22625/toc.htm which as usual for oracle documentation is excellent.

Last week i had preinstalled all the software ready for the actual migration today - the install of the software was all fine and didn't affect any of the 12.1.0.1 components which i was worried about. A completely new weblogic installation was put on automatically which made things easier than in old versions where you had to go and find the exact specific version yourself which caused no end of people problems previously (similarly for the jdk version).

The actual process went pretty well and the upgrade of the OMS part was actually pretty straightforward once i got to that part (the backups i manually took of all the directories took about an hour) - the total time for our OMS upgrade was maybe 2 hours and we have 20 servers being monitored with about 200 targets.

The only areas that gave me problems were the agents (both before and after the patch).

The problem before the patch was one of the agents had somehow got an old plugin version which was incompatible with 12.1.0.3 OM, everything i tried to get round this got me nowhere and i ended up completely deleting the problem host/agent to allow me to continue (overkill i know but i couldn't see any other solution at the time)

That got me past the prechecks and then i just clicked through the various steps and everything went surprisingly smoothly.

At this point the oms was upgraded and everything seemed to work - though i did get an odd error about compatibility mode - though that was easily fixed:





And here's the proof



Next for the agents....

For the ones using the 64 bit linux software they just worked no problem (well apart from one but I'll come back to that). For the others i just had to download the appropriate software (in offline mode for me as we have restrictions on internet access). These were then uploaded to the repository and deployed via the agent upgrade process - again with no issues (well other than some servers not having enough space in the AGENT_HOME and TMP directories - again easily fixed and the process is re-runnable).

This left me with just one agent that wouldn;t co-operate which just happened to be the one on the OMS server - it was complaining that "Oracle Home Property Missing". The explanation on this was somewhat confusing but i could see relative to other targets.xml files there was no reference to the ORACLE_HOME location of the agent software in the file. The fix just seemed to be saying run

emctl control agent runCollection <TARGET_NAME>:oracle_home oracle_home_config

Without really explaing how to get the TARGET_NAME. However in my case there was nothing at all in the file let alone just a missing ORACLE_HOME. After much puzzling and thinking about just creating an entry myself (i was put off by the TARGET_GUID which i couldn;t just make up) i discovered that the home had been autodiscovered by the agent itself just not promoted to be visible in the OMS/agent fully. To fix this i promoted it and waited a couple of minutes and it quite happily then upgraded the agent and plugins.

To promote - go to setup->add target->auto discovery results

then click on the agent based targets and find the entry for the oracle_home linked to the agent, promote that and it's resolved!

Now to see what's changed.....

The pie chart on the summary screen is different for a start.....

0 comments:

Post a Comment