Hi,
now for the answer.
Some very good suggestions all round - thanks for all contributions..
The box had been replaced with an identical box. Same kernel panic.
Yes, internet is provided by a 3G cellular.
The centos kernel had been upgraded to the latest available on yum.
The kicker was noticing at this point that sftp did not work and would
not connect.
Turns out the belkin router was faulty. A replacement was installed and
the customer is happy.
There have been no reported incidents since the new router was
installed, so I can only conclude
that it is all good.
The upgrade of the kernel happened at the same time as the router was
installed. Since the kernel panics
are sporadic, one cannot "know for sure" if it was the kernel upgrade or
the the router replaced that
stopped the panics. My view is that the older kernel version was not
coping with the huge number of
tcp fragments and retries. It would have been "leaking memory" or some
such thing until the box
died. Which is actually a very good suggestion - I could log and report
memory usage.
The suggestions to install "some-tool" are valid for home machines, but
not machines in the field.
Some googling reports that netstat reports a cumulative count of TCP
error messages, but
it is not clear which are the essential things to look at and use. I
suspect that some answers are valid
sometimes, and other answers are valid other times.
If someone has a favourite netstat command as a diagnostic tool to
report on the network link
I would appreciate very much.
================
One suggestion (off list was)
Probably environmental then: dirty power, magnetism, cleaner ( is there
a regular pattern to the crashes? ), script kiddies on the uplink or
somesuch.
However, as Craig mentioned, if it's still in that state and published
on the web, then security patches are essential...
My advice - look in /var/log and /var/crash - is still valid, and should
be the first step anyway.
======================
On 13/09/12 19:37, Christopher Sawtell wrote:
On 13 September 2012 18:47, Mark Beharrell <[email protected]
<mailto:[email protected]>> 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.
--
Derek J Smithies Ph.D.
Christchurch,
New Zealand
-- "How did you make it work??" "the usual, got everything right"
_______________________________________________
Linux-users mailing list
[email protected]
http://lists.canterbury.ac.nz/mailman/listinfo/linux-users
