<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=ISO-8859-1" http-equiv="Content-Type">
</head>
<body bgcolor="#ffffff" text="#000000">
ryan patterson wrote:
<blockquote
cite="mid:877e3bc60708170510m1d6a9231j6e09b83365b6741c@mail.gmail.com"
type="cite">
<meta http-equiv="Context-Type"
content="text/html; charset=ISO-8859-1">
<br>
<br>
<div><span>On 8/17/07, <b>Nick Morrott</b> <<a
moz-do-not-send="true" href="mailto:knowledgejunkie@gmail.com">knowledgejunkie@gmail.com</a>>
wrote:</span>
<blockquote>On 17/08/07, Jeff Bevis <<a moz-do-not-send="true"
href="mailto:jbevis@4access-comm.com">jbevis@4access-comm.com</a>>
wrote:<br>
<br>
> 2007-08-08 09:33:17.402 Unknown file transfer socket: 0 <br>
> BACKEND DEAD - RESTARTING<br>
<br>
As your crashes all seem to be relating to file transfers between<br>
hosts, I would start investigating here.<br>
<br>
--<br>
Nick<br>
</blockquote>
</div>
<br>
I think I have the same problem on my mythtv setup. I have a backend
(ubuntu 7.10 edgy) and a couple of networked frontends (ubuntu 8.04
feisty). The backend will sometimes crash while watching live TV.
When it crashes the backend log looks a lot like what Jeff Bevis
posted. I think it happens when one program ends and another begins.
I never use the "watch live TV" function so I have never actually seen
it crash. But my other users see it happen a lot. <br>
<br>
Jeff, Could you post your cron script that you wrote? Right now
sometimes my MythTV users will crash the backend and it will sit idle
until I get home and restart it. Having a script that will check and
restart the backend every five minuets would be great. <br>
</blockquote>
I don't have access to my home machine from work at the moment, but the
basic idea with it is like this: cron runs a bash script every five
minutes; the bash script executes "/etc/init.d/mythbackend status" and
greps the output for a unique portion of the dreaded <big>"</big><font
size="-1"><big><b>mythbackend dead</b>, but subsystem <b>locked" </b>message.
(if you look at the init.d/mythbackend script, you'll see that there
are a few possible messages that you might get returned). Upon finding
the backend to be dead (grep returns zero if it finds a match, nonzero
otherwise), it appends to the log file my little, uniquely-greppable
message, then executes "/etc/init.d/mythbackend restart" to get things
going again.<br>
<br>
I'll try to post my actual script and cron line when I get back home.
It's pretty trivial. Actually, since doing all that I've been looking
into monit as a replacement; you might want to consider that method of
monitoring since it is far more featured and configurable. Not that I
have even managed to set it up yet for myself <grin>.<br>
</big></font><br>
Thanks,<br>
-Jeff<br>
<br>
</body>
</html>