Thanks very much for this article. I just fixed a major fault on a production environment with your solution, avoiding an unneccessary database bounce and further interruption to service.

Andy
2nd line Support DBA at BT

The above from a 2-node prod RAC ;
INST_ID HANDLE GRANT_LEVEL REQUEST_LEVEL RESOURCE_NAME1 RESOURCE_NAME2 PID TRANSACTION_ID0 TRANSACTION_ID1 OWNER_NODE BLOCKED BLOCKER STATE
1 C0000001BB9A3B40 KJUSEREX KJUSEREX [0x19][0x2],[RS] 25,2,RS 1313 0 0 0 0 1 GRANTED
2 C0000001BB9A3C90 KJUSERNL KJUSEREX [0x19][0x2],[RS] 25,2,RS 20612 0 0 1 1 0 OPENING

both pointing to CKPT background process of node 1 blocking node2 ; what is the reason for this ?

Also ; is there a script to identify the session as well in the same query ?

Thanks.

Many thanks for the great article!

I’m trying your library cache lock example and looks like in 11.2 RAC the self-deadlock (when you compiles your procedure from inside the same procedure ) is now auto detected and the “ORA-04020: deadlock detected while trying to lock object” exception is thrown. Maybe because in RAC it’s a global resource and since 11.2 they added checks to detect this kind of deadlocks.

In 10.2 RAC and 11.2 single instance it works exactly like you explained – the session hangs trying to compile the procedure while executing it and x$kglpn shows two rows for the same session.

Without looking at details, I would guess that your developers are executing some stored object, then trying to recompile the object (may be, their “fix” didn’t work).
Please go through my blog and print the output of the scripts in the blog. May be, I can review details and help you.
If this is a deadlock, then I would need a deadlock trace file to understand the issue.

Thanks
Riyaj

Your “internals of locks” link is not accessible – can you kindly correct? Thank you for your really good articles.

Naresh