Oracle 12c as Metadata Repository

ray.wurlod · Post by **ray.wurlod** » Sun Mar 26, 2017 3:34 pm

Has anyone tried to use Oracle 12c as the metadata repositories' database and, if so, was it successful?

Please note that I HAVE read the system requirements documentation and do not need to be lectured on their contents.

chulett · Post by **chulett** » Sun Mar 26, 2017 5:00 pm

LOL... surely no one would dare.

We've just completed updating all of our instances to 12c, some went just fine, others... not so much. And with Informatica, some of the workflows were unaffected while our main nightly warehouse load takes twice as long now. Not a tragedy since it went from an hour to two hours but still... trying to track down the why of it. This after failing the first night due to a missing configuration item that the upgrader missed. Fun.

Much luck to you, Mr. Wurlod.

ray.wurlod · Post by **ray.wurlod** » Tue Mar 28, 2017 3:47 am

Might we know what the missing configuration item was?

chulett · Post by **chulett** » Tue Mar 28, 2017 7:56 am

I'll see what I can find out, see if "they" will share. Perhaps more of an issue if you upgrade an 11g instance to 12c rather than build one from scratch.

ray.wurlod · Post by **ray.wurlod** » Wed Mar 29, 2017 12:25 am

Just to be more specific, the requirement is for Oracle 12c Release 2.

chulett · Post by **chulett** » Fri Mar 31, 2017 12:45 pm

FYI - asked a couple of times now, not sure if they will actually come through for me or not.

Also, we continue to struggle with odd little issues that 12c introduced. No clue if/how any of this would affect your installation. Current fun we're having are "PGA Aggregate" errors, something specific to 12c:

ORA-04036: PGA memory used by the instance exceeds PGA_AGGREGATE_LIMIT

The limit is exceeded and Oracle kills off processes to "fix" the issue. Problem is two fold - if they are a job running SQL then the job fails, other times it is an internal Informatica process and rescheduling something fails. The former we can handle after notification, the latter snuck up on us as no failures are logged and no alerts go out. It's just that something that runs "every 15 minutes" runs then tries to update the repository to reschedule for the next 15 minute mark and that's killed by Oracle. Everything shows as "Succeeded" but it does not run again until someone notices the issue. Like say our external client. [sigh] I've had to modify our hourly monitoring script to also check for this specific situation. DBA still attempting to find a proper fix for us.

Hope yours goes better than ours.