On 13/09/12 10:46, Derek Smithies wrote:
Hi,
No activity for some time, so I thought I would kick things off with
a question.
Any suggestions (including shell commands) are welcome..
I have a problematic box on a customers site in the US. it is not
going to be easy to get physical access. SSH access is doable.
I have been writing software for a guy in the US - in which the server
maintains 20 concurrent TCP connections,
there will be as many UDP streams. Most of the UDP stream are inactive
at any given point in time. Usually - there are
no UDP flows - or 4 UDP flows. Sometimes 8 or 12 UDP streams.
The TCP connections are the control information to start/stop the
voice streams.
The UDP streams are voice streams.
This is a non standard variant on IAX2 - a better variant, but that
is another discussion.
This particular software is running on a centos 5.2 box. There are
hundreds of installations using this software in the states...
One customer has a cable modem (which is something like an ADSL box
without a NAT) and then a Belkin router (which is a NAT+wireless
access point+lots of ethernet ports for the lan side). The connection
to the public internet is via a 3G type link that has a mtu of 1000.
Not sure on the exact specifics of the internet connection - the
description made no sense to me.
The customer has reported kernel panics - many of them.
Memory checks? yes. Replaced all memory sticks with nice proper good
verified memory..
Program faults? This program is running fine on all other
installations - kernel panics have not been reported before.
CPU overload. No. loadavg is < 1, cpu busy percentage is 20%.
I am not installing mrtg (or similar) tool on this box.
ping tests are ok - you can ssh into this box on his site. Most of the
time, the box works fine and conveys all voice data,
all web traffic, all TCP commands just fine... except for these kernel
panics. Now, the sad part of the whole diagnostic
process is that I am not seeing the kernel panics in front of me. The
customer reports the kernel panics to the guy
I have been working for, who then reports them to me.
Thanks,
Derek.
First thoughts are 32/64 bit incompatibility (unlikely to cause a kernel
panic) / (yum) system updates not being done/out of sync / OOM.
Can you seen anything in /var/log or /var/crash? If you're lucky you might.
Alternatively, if you want to install munin-node on the server, I'll
monitor it for you???
Cheers,
Steve
_______________________________________________
Linux-users mailing list
[email protected]
http://lists.canterbury.ac.nz/mailman/listinfo/linux-users
