I started to use the Oracle 10g Grid at work, and have been configuring Agents for our various Oracle servers so I would have a central location to monitor them, as well as eventually attempt replication.
The BIGGEST problem I’ve had was trying to get the agents to talk to the 10g Grid Enterprise Manager server. I would do something like this:
> emctl upload
And get something like this:
“EMD upload error: uploadXMLFiles skipped :: OMS version not checked yet”
I’ve read so many different forums on how to handle this error, but none of them solved my problem.
What was going on was the hostname for the 10g grid upload read something like:
The funny thing was, if I went to the Repo URL on the actual server, I get a nice page that says it works, but if I tried it on the server with the agent installed, I wouldn’t get anything at all. After around 13 hours of fiddling around with this, looking at emagent.trc, listener.ora, emd.properties, trying emctl clearstate agent, securing the agent, unsecuring the agent, etc, I found out what was up:
For some reason that fully qualified domain name does not register with the other servers in the network. What I did was in my hosts file, I added a definition for the Repo URL and the IP it went to.
After that, I got:
> emctl upload
EMD upload completed successfully
You don’t know how happy I am right now.
Also, just an FYI: 10g x86 does not install well on a Windows x64 edition. What happens is the entire server hangs on a file in the %agent%/bin directory called fix_4198214.exe. During the installation phase of the agent (or the Grid), when that directory has that file copied over, just stop the installation (does not actually stop, but pauses it), and rename the file to something else. When the installer encounters the file, it’ll give an error message. Ignore it and you can happily proceed with the install