Message5752
Unfortunately, the bug is fairly hard to trigger. I couldn't reproduce it on my personal development system. You probably need: several threads, memory pressure to trigger garbage collection. At least this are the conditions on my pre-production system. I observed the deadlock twice within about 8-millions transactions.
I have attached the thread dumps of both cases. Common to both cases is, that the "RUNNABLE" thread consumes 100% CPU, doesn't perform any system calls and does not return from java.util.HashMap.get / java.util.HashMap.removeEntryForKey. Both methods traverse a linked list and this became cyclic. |
|
Date |
User |
Action |
Args |
2010-04-27 07:12:14 | akruis | set | messageid: <1272352334.64.0.201297201386.issue1602@psf.upfronthosting.co.za> |
2010-04-27 07:12:14 | akruis | set | recipients:
+ akruis, zyasoft |
2010-04-27 07:12:14 | akruis | link | issue1602 messages |
2010-04-27 07:12:13 | akruis | create | |
|