I ran into a problem while attempting to apply the April 2008 Critical Patch Update. We have a 4 node RAC cluster running 10g R1. I dropped the first node out of the cluster by shutting down the instance on the first node and stopping the node apps using srvctl. The OPatch utility was patched to the requird 1.0.0.0.57 and added to the path. I kicked off the opatch apply –local and it completed successfully.

The problem came up when I moved on to the next 2 nodes. Since the first 1 went so smoothly I dropped both 03 and 04 out of the pool and repeated the steps. They both applied without error until it got to the step where it started recording and verifying against the inventory:

OPatch detected dangling patch filemaps in the inventory. It will
clean up the inventory so that you can apply/rollback a patch successfully.

First, it will back up files that it's going to modify, then it will remove
bad entries in the filemap.

Verification of the patch failed
I found a metalink article entitled 1“How To Repair A Corrupted Inventory For RDBMS” which provides the script ‘checkinv.pl’ for diagnosing and repairing dangling patches. Running the script did identify the 6864078 as a dangling patch and remove it successfully. Unfortunately, OPatch still failed with the same error after reattempting the patch.

I ended up finding another article on metalink entitled 2“Updating the Inventory Is Failing During Patching.” The document suggests running opatch with OPATCH_DEBUG=TRUE. The output will provide checksums for comparing the files:

If everything output correctly the article suggests that the problem is not with the text formatted inventory but with the binary inventory being out of sync. To correct the issue we simply made a backup of the oraInventory\Content directory and deleted it. When OPatch ran it recreated the binary version and installed the patch successfully.

1. 1. https://metalink.oracle.com/metalink/plsql/ml2_bookmarks.addBookmark?p_t...
2. 2. https://metalink.oracle.com/metalink/plsql/ml2_bookmarks.addBookmark?p_t...