[mythtv-users] myth box crashed hard...

[Ray Olszewski]

Since a true "hard crash" clobbers everything (or, just possibly, 
everything except the kernel), it is a difficult problem to diagnose, And 
any onboard tool for diagnosing crashes would itself be hit by the crash 
... maybe too soon even to log what was happening just before the crash 
(depends on the timing of drive syncs).

So ... when I'm ancountering this sort of problem, I normally start an ssh 
connection well before the problem occurs and run "top" over it. When the 
test system crashes, the monitoring system still displays the last update 
of "top", which gives me some clues about what happened (as well as a close 
estimate of the time when it happened).

Also, you say "the network connection was also down". In  troubleshooting 
what looks like a hard crash, it is worth seeing if the system will respond 
to pings. Ping response is handled entirely inside the kernel, and 
sometimes it survives a failure that completely clobbers userspace, which 
helps you distinguish kernel-level and user-level failures.

Finally, if there is a kernel-level problem, the kernel will often post an 
OOPS message to STDERR. You may be lucky enough that this gets to the logs 
(I think this is sync-dependent). If not, it helps if you have a VT rather 
than X displaying on the system, since the display may survive the crash 
(depends on your hardware, I think)  and leave you with some information 
(not complete, since you can't scroll back, but at least the last 24 
lines)about what happened.

This is not a recipe; I don't know a recipe for this level of 
troubleshooting. Sorry I can't be of more help.

At 09:25 AM 9/11/2003 -0700, and wrote:
>My epia m10k, 2.4.20-8,atrpms myth0.11 box crashed hard last night and it 
>therefore forgot to record a program this morning. I had to hit the power 
>button as everything was completely frozen so I couldn't do a thing with 
>the keyboard and the network connection was also down, so I couldn't ssh 
>in either... I'd like to prevent it in the future, but have no idea how to 
>go about it.
>
>How do I track down the offending element?
>- Is there a tool to see when and what is doing the crashing?
>
>Sorry if this is a stupid linux question, but it is my hope that someone 
>else on the list have experiences that might help.... ;-)