Random System Crashes
 

Hi,

I'm still relatively new with Linux; but have been experiementing with it for the last 2 years and have grown extremely fond of it. I've built my own Linux HTPC Backend system running MythTV (I can post the specs if required) but I'm having an issue with the machine randomly crashing into a state whereby removing power is the only available option.

I can't track down a reason as to why this is happening and all the log entries end when the system goes down with no record of what happens. The screen still displays whatever was there before but mouse and keyboard appear disabled. Power button and reset button do nothing.

In my experience, this points to some sort of hardware failure. But with no evidence as to whats happening, I have no idea. This seems to have started occuring since the upgrade to FC6. In FC5 myth would often crash and there were numerous other issues with speed that made it tempting to move up to FC6.

Because the box suddenly crashes, there is no way for me to ssh or vnc into the thing to restart it and having to traipse all the way home to reset it isn't always an option - especially if it should be recording something.

I'm running Fedora Core 6 (2.6.20-1.2933), but I dont know what details you need to diagnose this kind of issue, so I dont know what to post - even a kernel panic would probably be more use!

500W PSU
AMD Athlon 64 3200+ (under clocked down to 1Ghz)
1024MB Ram (2x512mb)
2x250GB media drives
1x80GB os drive
2xHauppage Nova-T cards
1xDigiTV Card

Thanks for the help,
Chris

One trouble shooting step I can think of is remove one (2x512mb) RAM sticks at a time to see what happens, you might have a bad stick or it's loose as in chip creap.

Uhmm... The system runs in under 256mb most of the time anyway, so I'll try it for a couple of weeks either way. I'm leaning towards this being a graphics card issue, or some bug in FC6.

Because the bug is so random, its really hard to test. It's been nearly 2 weeks since the last time it happened, but before that it was happening every night if I left it running KTorrent, and only lasted 5 minutes if I ran VMWare.

I wonder if running the box in text mode might remove the possibility of it being related to some dodgy ATI drivers... I really dont like the card thats in there (an X600 ATI).

I can survive on the terminal for a while, its only the backend so it doesn't need to do any video output.

Whats the best way to remove the drivers and make it run in some form of crappy BIOS standard with no hardware acceleration?

Mine did that when the temperature sensors on my graphics card stopped working. In fact, the system hadn't hung, just the display adaptor. SSH sessions still worked.
I presume you've not got another graphics card to run in place of the current one to test?

run as root will drop you into text-only.

To make it default, open up /etc/inittab (again as root) in your favourite text editor and you want the line that reads
to read
i.e. you replace the 5 (runlevel 5 is multi-user and graphical) with 3 (multi-user text-only).

If you still need X, try the opensource radeon drivers instead of the ati drivers... or even the crappy VESA drivers... just change the appropriate "Driver" line in xorg.conf to "radeon" or "vesa".

If I'm only using vnc 90% of the time, and any other time I go on the box I can use console directly... do I need to load anything other than VESA? - or does the gfx card still do work over vnc ?

I was under the impression vnc was completely software based, but I'd just like to clarify.

I've dropped the system down to runlevel 3, and have been running X through vnc... no problems yet.

I've also sat at the box and started X, played with it for a couple of days and then closed it down also without any issues.

If it happens again, I'm going to consider switching distribution to either flat debian, or xubuntu/kubuntu. Then I can elimiate the issue being with fedora... although it wont tell me if it's myth or not...

I have a bad feeling its actually one of the drivers for the TV tuners or the tuners themselves.

Chris

I may end up buying a new gfx card for my workstation, so this one may inherit a PCI-Ex Geforce 6200... it's been stable for the entire time I've been using it on my workstation so it may be a viable option.

I wish you could run these things without graphics cards, hehe.

Cheers,
Chris