Java VM locking on highly multithreaded Java apps
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
glibc (Ubuntu) |
Fix Released
|
Critical
|
Jeff Bailey |
Bug Description
Kevin A. Burton on debian-glibc:
We're running a java application server thats highly multithreaded and
we're noticing that the VM is locking hard after about 30 minutes
when I increase the thread count.
This is still a lot of threads (600 or so) but there's no reason the
entire process should lock. It also won't respond to a kill -9...
http://
The problem is fixed upstream, but because of the Sarge freeze
no patches were accepted for this glibc problem.
Two patches from FC1 were supposed by Mark Spitzer in:
http://
I supplied these patches and run the test from Redhat's
Bugzilla for RHEL 3 about the same problem
https:/
But the two patches didn't fix the thread locking. So I extracted
an RHEL 3 glibc SRPM and by comparing the dates from Redhats bugzilla
to the date of the patches in the glibc SRPM I identified a third
patch which could solve the problem.
After applying this third patch the test program works. (No more
thread locking)
I've been running with these patches for three weeks now on a production
server without any JVM lockups.
Please consider applying these patches - a locking Java VM is not much
fun when you need to run big J2EE applications.
BTW: RHEL3 2.3.2 glibc has much more important (as far as I can tell)
fixes than the Debian version. It would be wise to track this version
more closely.
Thanks for Your work on Ubuntu,
Gerald Hochegger
http://
*** Bug 14410 has been marked as a duplicate of this bug. ***