[ofa-general] [Fwd: [Error] Asynchronous Thread]

[Dotan Barak]

Yann K. wrote:
>
>
> ------------------------------------------------------------------------
>
> Subject:
> [Error] Asynchronous Thread
> From:
> "Yann K." <yann.kalemkarian at bull.net>
> Date:
> Thu, 21 Jun 2007 16:50:59 +0200
> To:
> ewg-bounces at lists.openfabrics.org
>
> To:
> ewg-bounces at lists.openfabrics.org
>
>
> Hello everybody,
>
> I have a problem making a diagnostic on those kind of errors, which 
> happen at the same time :
>
> At the mpi level :
>
>        case IBV_EVENT_SRQ_ERR:
>            ibv_error_abort(GEN_EXIT_ERR, "MPI Gen2 Async Special Event 
> thread : Got FATAL event %d\n",
>                            event.event_type);
>
> At the kernel level :
>
> Jun 21 11:17:55 s_kernel at platine866 kernel: ib_mthca 0000:07:00.0: CQ
>> overrun on CQN c2009c
It seems that you got CQ overrun which means that more completions that 
the CQ size were created.
You can solve this by creating a bigger CQ or use more than one CQ...

 (i don't really understand why you sent the code from the MPI which 
handles SRQ error).

thanks
Dotan