I have a problem with IDS 9.21.UC2, OPENSERVER 5.0.7:
Now and then the machine is crashing with the online.log-message:
Assert Failed: Mutex Holder Value is not Null
Who: Session(7592, ...(User)..., 1348, -1166998708)
Thread(8152, sqlexec, b713c634, 1)
File: mt.c Line: 7049
stack trace for pid 459 written to /tmp/af.23c0ebef
See Also: /tmp/af.23c0ebef, shmem.23c0ebef.0
Here's the stack to one of the crashes written in the af-File:
I already read that this could be because the instance is having problems freeing the cache and that it could help to run oncheck -cI resp. -cD on the
tables that were accessed to in the moment on the crash. Also it should help
to rebuild the indexes of the tables affected.
Problem is, that this version is out-of-support at IBM and there are no new versions available for this platform, so I won't be able to have the SHM-Dump analysed.
Anybody got a clue what causes this problem or how it could be fixed?
Look for on the documentation I found some bugs with relation the message "Mutex Holder Value is not Null", but all are relation with PDQ, SBLOB, Replication.
do you use anything? PDQ, SBLOB, Replication?
Thanks for your reply, but I'm not sure if disabling the statement cache is an option because this would cause a performance problem, would it not?
These crashes just occur very infrequently, one time 4 months ago, but two times last week. Perhaps it would take months before and if this happens again.
There are always other tables involved and the statements were always different.
As for the semaphores, PDQ, SBLOB, Replication, I am not sure and will check it out.
Where can I find the doumentation you were talking about? Something related to IDS 9.21 and not to 9.40 perhaps ?
I know this version is out of support, but switching to another would mean to switch to another OS as well.