8.0 -> 8.10 Upgrade: Spec Merge Memory Violation

[jde_cnc]

List,

More fun with upgrades:

On an 8.0 to 8.10 upgrade, the Spec Merge is failing with a Memory Violation after about 10 minutes. Message is as follows:

"Business Function AddCumObject at Line 34 for Event in Section Merge Specs has caused a memory violation."

Oracle is largely at a loss right now, but it looks like this may also be related to the use of OLEDB. All legacy database connections are ODBC, as are the new connections. However, I noticed that the connections in the Upgrade Plan from the 8.10 system down to the 8.0 system were created by default as OLEDB. Solution 200949255 says that OLEDB is quite error-prone on this platform (Win2k3/SQL), so I am trying to verify that this is causing the problem.

Any thoughts?

 

[jstooge]

I'm surprised that Oracle didn't tell you to definitively change the connections. The connection issue started out for just CO's, but has been expanded to include all. In fact, in 8.11, the default is back to ODBC. Anyway, yes, I'd set everything to ODBC and retry.

 

[jde_cnc]

Yeah, as of my second call with them, that is now their advice. I am doing so, and hoping that this clears the issue. Thanks.

 

[jde_cnc]

Replaced all OLEDB data sources with valid ODBC connections, but no dice. The same memory violation is continuing. Is there a viable way to find out exactly what object merge is causing the error? The merge runs for a while, and a restart of the process creates a prompt that some spec merges have already run, so it looks like a specific object is causing it.

Thanks in advance.

 

[jstooge]

2 places most probably, in the plan you are running, you can go in and look at where it left off, or you should be able to use P98700.

 

[shawnscanlon]

You may want to verify that you do not have any objects with a name longer than 10 characters. This often times will cause memory violations and problems with package builds as well.

 

[jde_cnc]

I will take a hard look at that now. Looks like many of the spec merges completed successfully, so I know that the whole process did not fail.