Re: OEM 10g Greed Control

  • From: Robyn <robyn.sands@xxxxxxxxx>
  • To: Rich.Jesse@xxxxxxxxxxxxxxxxx
  • Date: Thu, 23 Jun 2005 14:40:57 -0400

Grid world is looking better today; all the targets are online and 
reporting. Here's an update:

Dynamic discovery worked perfectly on one server. All the others (4 thus 
> far) have had to be manually added. Not bad in testing, but in a large 
> organization, this would take forever. One of them won't talk back to the 
> management db at all; everything on that server stays pending forever. 
> (planning on trying to fix that one when I finish this note.)
> 

I shutdown all the services, ran 'emctl clearstate' and the agents and 
started wading through the config files. For the one server that wasn't 
reporting correctly, I cleared out the following files per note 286271.1:

Agent $OH/sysman/emd/upload/*.*
Agent $OH/sysman/emd/state/*.*
Agent $OH/sysman/emd/collections/*.*
Agent $OH/sysman/emd/agntstmp.xml
Agent $OH/sysman/emd/protocol.ini
Agent $OH/sysman/emd/blackouts.xml
Agent $OH/sysman/emd/lastupload.xml

This note referenced running different versions of the agent to the control 
center. I verified that all agents were on the same version, but the 
symptoms of the problem were similar enough that I decided to try clearing 
the agent for one server anyway. Oddly enough, it actually worked, the agent 
came on line and started reporting for all targets on the server - except 
the host itself. (How can the agent report itself as running on a host, but 
not know about the host it's running on?) I manually added the host line to 
the targets.xml file, and voila! everything is up. Plus, the databases were 
dynamically discovered this time, even though they had been discovered and 
removed several times.


8) The Agent's processes regularly cause forty SYS/PUBLIC objects to
> > become invalid, flagging a policy violation in EM. The workaround is to 
> > drop that policy.
> 
> 
> Seen this one too. I recompiled the procedures with utlrp.sql and thus 
> far, the procedures stayed valid, but it's only been 24 hours ...
> 

All the sys procs are staying valid after I rerun the utlrp.sql script. I've 
been revoking some of the execute privs from public as I bring the servers 
on line; does the revoking of these packages cause the procedures to go 
invalid or is it OEM? I did have to explicitly grant execute to sysman that 
I had revoked from public. Need to experiment more and see when it breaks.

I'm feeling pretty brave now that all my HP-UX test servers are agentized. 
Time to add an MS server into the mix just to get a little excitement back 
in my day ...

Robyn

Other related posts: