http://damir-vadas.blogspot.com/2009/11/deadlock-deadlock-how-to-monitor.html

Cheers!

Sorry, apparently I’m useless. I can no longer query the SR, so, I think it’s been closed too long and aged out of MetaLink. As to the deadlock issue, no, I don’t think I filed a bug. I just created the missing indexes and moved on.

-Mark

Which bug did you hit when you ran the catalog upgrade? Did you log the deadlock issue as a bug?

We are doing a lot 11g testing now and I am trying to document these issues in our upgrade documentation.

Thanks!

Dallas

yes, from what my collegue Chris has investigated, the issue is 10.2.0.3 rdbms backing up with 11.1.0.6 catalog (fine for 11.1.0.6 rdbms backing up with 11.1.0.6 catalog. So you are fine with the first step but indeed, it is advisable to wait and not do the second step for now.

see you,
eric

Interesting, I was not aware of this bug. Can you clarify? It seems that this bug affects only 10.2 (or lower?) databases being backed up w/ 11g catalog? The problem does not occur on 11g database backed up using 11g catalog? If so, I think I’ll be ok, for the time being. So far, we are maintaining the old 10g catalog as well as the new 11g catalog, and only using the 11g catalog for backing up the 11g databases. The plan was to start migrating the 10.2 and lower databases over to the 11g catalog. But, after learning about this bug, I think we’ll be holdling off on that, at least till a patch is available.

Thanks Eric!

-Mark

thank you for starting this blog. Already in my Google Reader favorites.

We have hit another issue with RMAN and 11.1.0.6 (see https://metalink.oracle.com/metalink/plsql/ml2_documents.showDocument?p_database_id=NOT&p_id=577600.1)
We had to go back to 10.2 for the RMAN repository (all work done by my colleague).

see you,
eric

The RMAN codebase and table structures are the same

So, it sounds like the performance of some bit or bits of code that 10.2.0.3 RMAN runs has degraded in 11g compared to 10.2.03?

In which case, something that you might expect from an upgrade – some code gets better, some stays the same, some gets worse. And as it’s internal oracle code, would that be something that Oracle would have tested as part of 11g development – 10.2.0.3 RMAN on Oracle 11g?

I have several trace files, all indicate the same type of deadlock graph, all holding TM enqueue in ‘SX’ mode, and waiting on ‘SSX’ mode. There are at least two sessions involved in each deadlock, these are not self-deadlocks. Also, adding the indexes that were missing has clearly solved the problem.

-Mark