For the past 1.5 years, starting at Oracle 8.1.7.3.2 and now 8.1.7.4.10 on
windows; we have encountered a problem wherein we encounter a runaway thread
during patching of Oracle Applications 11i. This thread is not know to
v$session and orakill will not terminate it. The only solution has been
stop the database and service.
Recently, following implementation of HRMS F and its prerequisites
(including FND G), the runaway thread problem mutated to a major thread
leakage. In our dev/test instances, with the above patches installed, we
would hit 1000 threads assigned to oracle.exe per task manager!!!!
After a long ODC session with support and continuously monitoring of the
threads, it appeared that the "extra" thread created for each connection
was not always being terminated. You get two threads per connect of which
one appears to be for DCD. Friday, we turned off ODC in test and appear
stable. We are now trying in dev to confirm this may be a solution.
If so, we will turn it off in production and implement the patches.
Does anyone have similar problems with DCD. We had seen some bugs on
this but they were with 8.1.7.2.x. During our ODC session, we asked about
known bugs and there were none.
Jeffrey Beckstrom
Database Administrator Greater Cleveland Regional Transit Authority 1240 W. 6th Street Cleveland, Ohio 44113 (216) 781-4204 |