[Wylug-discuss] System Crash

[Mark P. Conmy]

On Fri, 6 Jul 2007, Chris Brown wrote:
> On 06/07/07, Jim Jackson <jj at franjam.org.uk> wrote:
>> On Fri, 6 Jul 2007, Chris Brown wrote:
>> 
>> > Hi Mark,
>> >
>> > On 06/07/07, Mark Spink <dmspink at aria.uklinux.net> wrote:
>> > >
>> > > Hi
>> > >
>> > > I have a co-location server thats has been running without a problem
>> > > for over 3 years. This morning we had our first problem in ages, we
>> > > could not connect to the web server or ssh, however we could still
>> > > ping the server. Diagnostics is not my main area of expertise and if
>> > > anyone could tell me where or how-to diagnose what happened I'd be
>> > > grateful, perhaps even a few pints grateful!
>> >
>> > Sorry to hear of your issues. So networking is still up but ssh and web
>> > server are not responding.
>> 
>> Being able to ping is not a great diagnostic. I've seen servers where
>> nothing userland was working but still responded to pings - ICMP is a
>> kernel function.
>
> Do you have any alternative suggestions? Mark admitted he is not a great
> diagnostician. He has asked for help, not criticism.
> 
>> > Can you browse to the web server by ip address or login to ssh via
>> > ip as opposed to domain?
>> 
>> errr.... the problem's with his server, not his client.
> 
> How do you know that? You are making assumptions based on very little
> information. I would respectfully ask that you - and anyone else
> responding to this thread in the future - assist Mark or not respond
> at all.

Chris,

Take a deep breath and read it again imagining the response being sent
by someone with a sympathetic look on their face.  I suspect Jim wasn't
criticising (I didn't read it that way) but simply explaining that
little can be divined through the presence (or, these days, absence) of
ping (specified by the ICMP protocol) responses.

As to the problem itself, there is (sadly) little that can be done
remotely unless there is a way to get to the console (IPMI*, IP-enabled
KVM or similar) or the userland services start playing ball again.

In my experience, however, this is likely to be a fatal oops and require
a reboot.  It may or may not be hardware inspired, but there really
isn't enough information to know based on what's been said so far.

Regards,

Mark

* IPMI isn't likely to be an option given the vintage - it existed, but
   it was an early version that wasn't often configured.