Best way to troubleshoot intermittant lockups on F12

[Mark Goldberg]

On Fri, Jan 22, 2010 at 10:26 AM, Marko Vojinovic wrote:
> On Friday 22 January 2010 16:07:10 Mark Goldberg wrote:
>> I have an F12 installation that experiences intermittent lockups,
>> usually at times of heavy usage.
>> The system just locks up solid, keyboard and mouse are non responsive.
>
> Telnet? Why telnet?

It is easy to telnet in from a local machine and administer it from there.
>
> I would check for the following:
> * ssh from the remote machine --- this checks the extent of lockup. It may
> well be just X locking up, while kernel is still alive and well. That would be
> the issue with graphics hardware/drivers.

I think the telneting in is equivalent. It is not completely dead, but
can't even
finish a login.

> * reproduce the problem with a LiveCD, prefferably different distro --- this
> checks for hardware vs. software fault. If for ex. Knoppix locks up too, it's
> definitely a hardware issue, independent on the OS in use.

It is a new Motherboard and I imaged a disk from a backup with Knoppix
6.2 with no
problems. That does not prove anything because F12 can run for days with no
problems and then crash three times in a day.

> * if it's a hardware issue --- upgrade power supply, downclock the processor,
> change the motherboard, remove components from the box one by one until you
> find the culprit.

This is a Mythtv box, all the slots are full of tuners, SATA
controllers, Video, etc.
I know any of them could be the culprit. The only thing being used early in the
morning when it crashes is the cpu, hard disk and network during h.264 encoding.

I've done some of this. I ran several processor stress tests and it can go to
15% overclock before it starts to have problems.

> * if it's a Fedora issue --- boot with nomodeset kernel option, disable ACPI,
> change graphics drivers, boot older kernel, try to narrow down the app that
> triggers the lockup...

It has happened with several F12 kernels. I'll try some of the other
options you listed.

Is there anything better than remote rsyslog messages? When it
crashes, nothing comes
across. I believe it cannot do Disk I/O when it crashes, so it hangs,
but I've tested all the
drives and SMART shows no errors.

Thanks,

Mark