How to debug a remote server when it hangs???

Submitted by bret on Mon, 07/23/2007 - 09:21.

Here's one for you guys:

I installed a linux samba fileserver in Sitka about a year and a half ago, and it's been hanging (since Xmas), and I can't figure out why. It just stops responding. The times between hangs are not the same. It'll run for weeks and weeks, then hang, then 2 days, then hang, then a week and a half, then hang. I've disabled all the junk on there that isn't necessary for what I need it to do (postfix, xwindows, etc etc). How do you debug something you can't see?

I've been trying to get a monitor/keyboard hooked up to it so I can step through someone while it's hung, but for various reasons that hasn't been easy (mostly employee impatience). They learned that if they cycle power, it'll come back up and they'll be on their way, so that's what they do.

There's nothing in the logs that I can see that gives me any clues. You guys have any tips?

Linux version 2.6.15 (Debian 1:3.3.5-13)
512 MB RAM
1GHz AMD Athlon
Samba v2.0.23c-1

sounds like hardware to me. = w

Submitted by wax on Wed, 07/25/2007 - 19:45.

sounds like hardware to me.

=
w

Might want to check the temp, I think the lmsensors package does

Submitted by ryan on Mon, 07/23/2007 - 22:23.

Might want to check the temp, I think the lmsensors package does the job depending on the hardware. I worked on a dual CPU server the other day, one of the CPU fans had stopped turning, but the machine still "worked".

I'm dealing with a similar situation currently in a remote town.

Submitted by ryan on Tue, 07/24/2007 - 07:47.

I'm dealing with a similar situation currently in a remote town. The server comes and goes as it pleases. The first thing I did was install a VM on a spare workstation on the network, setup the same OS, and rsync the var, home, etc dirs over so I could get an operable/interim replacement while the main server was troubleshot (or possibly just shot).

And if you're having a hard time convincing folks it's time to r

Submitted by ryan on Tue, 07/24/2007 - 07:48.

And if you're having a hard time convincing folks it's time to replace their server, here's a handy command:

shutdown -h 0

two words - site visit

Submitted by colin on Wed, 07/25/2007 - 19:31.

two words - site visit

or these two : send colin

Submitted by random yacko (not verified) on Mon, 07/30/2007 - 07:58.

or these two : send colin

Post new comment

The content of this field is kept private and will not be shown publicly.