Suspended sessions in 10g
Date: Tue, 30 Dec 2008 09:53:14 -0600
We recently upgraded a test database from 220.127.116.11 to 10.2.0.4 in preparation for the production upgrade tomorrow. People were testing an application and some reports. Suddenly everyone reported that their session was "frozen". Some checking in the alert log revealed that there were many suspended sessions. They reported a datafile couldn't extend. But that datafile had plenty of space. So we checked the temp tablespace which was still stored as a data file (not a true tempfile) and dictionary managed. The report session that had hogged the temp space was killed and everything quickly returned to normal.
Has anyone seen a spurious message like this? We have converted the temp tablespace on the test database to a locally managed tempfile, but there won't be time to replicate the situation before the production conversion.
Our plan for production is to convert the temp tablespace to a locally managed tempfile as part of the upgrade. And create a separate temp tablespace for the report user.
Does anyone have any other suggestions?