[Cbe-oss-dev] unpredictable shutdowns

[Geoff Levand]

On 06/18/2009 05:26 AM, Akos Marton wrote:
> Fedora Core 9 is installed on one of my SONY PS3s. This machine
> shutdowns after unpredictable uptimes (after 4-7 days, but the last
> was 12 hour). 

Swap the hard disk with a one of your 'good' ps3s and see if the
behavior persists.

I haven't got any idea why it's happening. It is
> embarrassing, because I am working on these machines with remote
> shell. This machine situated in a rack, but I think there isn't high
> temperature, or may?

The ps3's lv1 hypervisor can report thermal events to the guest OS.
The linux kernel requests these events and will print a message in
the kernel log: 'PS3 Thermal Alert Zone XXX'.  I have never seen
one of these messages though.  See PS3_SM_EVENT_THERMAL_ALERT in
the kernel source file drivers/ps3/ps3-sys-manager.c

If the machine gets too hot the lv1 system manager should shut
it down.  I have never seen happen this either though.

> Have you got any idea, how could I debug or see a posted log (syslog,
> kernel log, etc.) about this problem?

Log in from another machine using ssh and do something like this:

  tail -f /var/log/messages

and/or

  killall rsyslogd
  cat /proc/kmsg

-Geoff