dtrace discuss - Jan 2007 - Diagnosing into server crash

Hi all,

can anyone point me to a document that descirbes how to trace into an OS 
crash under Solaris 10? My hope is to get target-oriented description to
figure out the system call and the process/thread that lead to the crash.

Regards,
Andreas

Andreas.Haas at Sun.COM wrote:
> Hi all,
> 
> can anyone point me to a document that descirbes how to trace into an OS 
> crash under Solaris 10? My hope is to get target-oriented description to
> figure out the system call and the process/thread that lead to the crash.
mdb''s $C will tell you that stack of the panicing thread ...
that''s a good start. If you
have access to SCAT, "panic" is a good start. the rest depends :-)

HTH
-- 
Michael Schuster
Sun Microsystems, Inc.

To use DTrace to analyze a crash dump, you need to have been running 
DTrace when the system crashed.  Take a look at chapter 37, Postmortem 
Tracing, in the Solaris Dynamic Tracing Guide.  The basic idea is that 
you can trace into a single ring buffer (wraps-around), and then after 
the crash, find the buffer and dump the trace output for whatever length 
ring buffer was allocated.

Chip

Andreas.Haas at Sun.COM wrote:
> Hi all,
>
> can anyone point me to a document that descirbes how to trace into an 
> OS crash under Solaris 10? My hope is to get target-oriented 
> description to
> figure out the system call and the process/thread that lead to the crash.
>
> Regards,
> Andreas
> _______________________________________________
> dtrace-discuss mailing list
> dtrace-discuss at opensolaris.org

Thanks to everyone for the replies!

Andreas

On Wed, 17 Jan 2007, Chip Bennett wrote:
> To use DTrace to analyze a crash dump, you need to have been running DTrace
> when the system crashed.  Take a look at chapter 37, Postmortem Tracing, in
> the Solaris Dynamic Tracing Guide.  The basic idea is that you can trace
into
> a single ring buffer (wraps-around), and then after the crash, find the 
> buffer and dump the trace output for whatever length ring buffer was 
> allocated.
>
> Chip
>
> Andreas.Haas at Sun.COM wrote:
>> Hi all,
>> 
>> can anyone point me to a document that descirbes how to trace into an
OS
>> crash under Solaris 10? My hope is to get target-oriented description
to
>> figure out the system call and the process/thread that lead to the
crash.
>> 
>> Regards,
>> Andreas
>> _______________________________________________
>> dtrace-discuss mailing list
>> dtrace-discuss at opensolaris.org
>
>