[ofa-general] Log output upon death

[Eitan Zahavi]

Hi Frank,

>From your description it is unclear if it is the ibmssh (the shell that
interprets the RunSimTest code)
Or OpenSM has crashed. The best way to debug such issues (sudden death)
is to compile the executables (both opensm and ibmssh)
with debug info (by adding -ggdb to CFLAGS or better configure
--enable-debug) and then allow the system to create core file
(in bash use: ulimit -c unlimitted; in tcsh limit core unlimit).

Then you will get a core dump file.
Yo ushould try to open it in gdb and it will tell you what executable
generated the core.

Then you start gdb with the correct executable and core file and use the
"where" command to debug.
You can switch between threads by using the thread command.

If you want me to have a look at the failure you can send me the "input"
files you use (topo file and ibnl directory).

Eitan

Eitan Zahavi
Senior Engineering Director, Software Architect
Mellanox Technologies LTD
Tel:+972-4-9097208
Fax:+972-4-9593245
P.O. Box 586 Yokneam 20692 ISRAEL

 

> -----Original Message-----
> From: general-bounces at lists.openfabrics.org 
> [mailto:general-bounces at lists.openfabrics.org] On Behalf Of 
> Frank Olaf Sem-Jacobsen
> Sent: Sunday, June 03, 2007 8:29 PM
> To: general at lists.openfabrics.org
> Subject: [ofa-general] Log output upon death
> 
> Time for my second naive question (too bad the archives do 
> not have any search function).
> 
> Much as expected RunSimTest dies for an unknown reason while 
> routing my topology, and I am attempting to debug by adding 
> various debug log entries.  However, as things seem to be 
> threaded (?) there does not seem to be any direct 
> relationship between where the application fails and where 
> the log output stops, the log usually stops abruptly in the 
> middle of a line.  Also, the log entries stop in various 
> parts of the code instead of the same place each time (I 
> could though have many errors ;) ).
> 
> Is there a possible way to synchronise this such that the log 
> file will reflect the last log entry by opensm before it 
> dies?  Are there any other ingenious ways of debugging the 
> route building function?
> 
> As always, any help is greatly appreciated.
> --
> Frank Olaf Sem-Jacobsen
> _______________________________________________
> general mailing list
> general at lists.openfabrics.org
> http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general
> 
> To unsubscribe, please visit 
> http://openib.org/mailman/listinfo/openib-general
>