Description of problem:
We are seeing a problem where a thread in our application starts consuming all
available cpu resources on the host. Our application is based on java and we are
using Sun Microsystems jre release 1.4.0_00. The problem occurs within the
java.lang.UnixProcessReaper module which runs in it's own thread and seems to
receive SIGSEGV's in a loop (I straced it's process id (it happens within
libc.o)). This is what seems to consume the cpu.
Our application is fairly simple - it collects host status info by issuing linux
commands like df etc. through java and reports the results back to a server
(tcp/ip). In parallel we have been stressing the system by executing the
Cerberus test suite.
We are providing different release of Redhat Linux precanned for our customers
based on Redhat 7.1, (7.2), 7.3 and AS2.1. Interestingly we only see this
problem on AS2.1 - not on 7.1, 7.3 (we haven't done as much testing on 7.2 as
with the other 3 releases)
Version-Release number of selected component (if applicable):
Multiprocessor systems - 2 and 4 cpu machines (32 bit Intel)
Run our client.
Run Cerberus testsuite.
Steps to Reproduce:
If you have an idea how to further debug it - it would be much appreciated we
will be able to leave a test system that is in the described state up and
running for a while.
Does this still happen after the QU2 update, released at the end of May, which
contained the e.24 kernel?
It's been over two weeks since I asked for more information. Thus, I am assuming
that the errata did fix the problem, and closing it. If this is not the case,