A big part of what DBAs do is troubleshoot. And a good DBA has a tenacious attitude toward troubleshooting. Also, he or she relentlessly applies cold hard logic to a problem to arrive at a solution. Hoping and wishing for a solution doesn’t cut it.
So when do you choose to let go of a problem, leaving it unsolved? I recently let go of a problem for which I know there is an explanation, but I don’t know just what.
I installed Oracle database 18.104.22.168.0 on a Solaris 10 sparc machine. Worked fine. I applied patch 22.214.171.124.0. The OUI gui said that worked fine. I did not check the inventory after the install. Then I applied several recommended patches for 11g that get installed with the OPatch tool. And OPatch came back and said the patches were not needed! How can this be, I thought. I opened a ticket with Oracle. Then I looked at the inventory and saw that the 126.96.36.199.0 installation did not show up in the inventory. However, 188.8.131.52.0 was definitely installed. The banners for sql*plus, exp, expdp and v$version all said so, as did dba_registry.
Oracle support felt that something had overwritten the comps.xml file in the oraInventory/ContentsXML directory. It didn’t look that way to me. There were some suspicous entries in the 184.108.40.206.0 oui log file, but nothing conclusive. Oracle support suggested that I install into a different home and try again.
So maybe it was a rogue file restore that overwrote comps.xml. Or maybe there is some strange bug in OUI that Oracle hasn’t hit upon yet. But I have filed this one under Unexplained Mysteries. It’s not worth my time to dwell on it further. I’ve got to keep my 11g research going!