I searched around the Net about the "thread tries to lock a nonlocked object" bug, and found a discussion about such an issue:
http://forum.java.sun.com/thread.jspa?threadID=343023&messageID=2942637 According to it, this might actually be a bug in glibc and nptl library. Specifically, there appears to be bug in GLIBC 2.3.2 and NPTL 0.60 which causes pthread_cond_broadcast() to be lost in some circumstances. According to http://www.blackdown.org/java-linux/java-linux at java.blackdown.org/java-linux-msg00089.html this can be tested wit gdb: + find out java process id + use gdb to attach to the process (gdb - <pid>) + quit gdb If things get back to normal after a few seconds, it was caused by this bug. Here's more info possibly related to the matter: http://www.mail-archive.com/debian-glibc at lists.debian.org/msg10837.html Of course, none of this helps any to solve the matter, but at least it would be nice to know what causes it... So, can anyone test this ? I can't, I don't know enough about Java debugging :(. Maybe we could have the wrapper start gdb, attach it to the running Java process, and quit every minute or so ?-)
