Description of problem: On my x440 (with 8 procs and 16gigs ht on) if
just run "javac" (not doing any work) I have javac processes still
running if I look on top. These processes are eating as much cpu time
as they can and they don't go way unless I explicitly kill -9 them.
The more I run javac the more zombies I get. This is not a good thing.
I did a little stracing and found the following.
if I strace the initial process all the rt_sigsuspend calls
look like this
rt_sigsuspend([USR1] <unfinished ...>
--- SIGRT_0 (Real-time signal 0) ---
<... rt_sigsuspend resumed> ) = -1 EINTR (Interrupted system call)
and if I strace one of the running zombies I see
--- SIGSEGV (Segmentation fault) ---
over and over again. They are stuck is some loop.
Version-Release number of selected component (if applicable):
I am runnig 2.4.9-e.12summit.
How reproducible: For me very easy
Steps to Reproduce:
1. type javac at a console
2. repeate step 1 as much as you want
zombie javac processes remain drasticaly degrading performance.
no extra proceses.
Additional info: Feel free to contact me for more debug info.
We have found that this bug has been resolved by IBM Java2 version 1.4.0.
Will be resolved in RHEL 3 which will bundle IBM JDKs 1.4.1.