All,

I realized that aborting a job omp.x < st.in will crash the kernel at 
the front end.
But not at the compute node.

If you let the job finished, it would not panic. Reproducible everytime.
Time to debug.

I have my suspicions.. will hold them...

Thanks,
Robin


Jaime E. Combariza wrote:

>please forward to Joe Simpson, I do not have his email handy.
>
>from unixgen.muohio.edu  ( ~ 4 sec)
>
>$
>$ ping -s foyer2.osc.edu
>PING foyer2.osc.edu: 56 data bytes
>64 bytes from foyer2.osc.edu (192.148.248.53): icmp_seq=0. time=4057. ms
>64 bytes from foyer2.osc.edu (192.148.248.53): icmp_seq=1. time=4137. ms
>64 bytes from foyer2.osc.edu (192.148.248.53): icmp_seq=2. time=4147. ms
>
>
>From redhawk (similar results)
><mulnx31|combarje> % ping foyer2.osc.edu   (~ 3sec)
>PING foyer2.osc.edu (192.148.248.53) 56(84) bytes of data.
>64 bytes from foyer2.osc.edu (192.148.248.53): icmp_seq=0 ttl=244
>time=2990 ms
>64 bytes from foyer2.osc.edu (192.148.248.53): icmp_seq=1 ttl=244
>time=2968 ms
>
>
>From a machine at Arizona:  a lot faster??????????
>
> <dyer.rc.arizona.edu|Colombia|jaimeco> % ping -s foyer2.osc.edu
>PING foyer2.osc.edu: 56 data bytes
>64 bytes from foyer2.osc.edu (192.148.248.53): icmp_seq=0. time=45. ms
>64 bytes from foyer2.osc.edu (192.148.248.53): icmp_seq=1. time=44. ms
>
>
>
>
>
>
>
>  
>
>>David Woods wrote:
>>
>>    
>>
>>>I'm wondering if we are seeing the lag to access the network license at
>>>OSC.
>>>
>>>I tried the 8.1 compilers, which aren't on the Ibrix file system, and it
>>>is also slow.  When I ping foyer2.osc.edu, which is their license server,
>>>the time lag is almost 4 seconds.  If this happens a couple of times when
>>>the compiler is used, this could explain why the compilers seem so slow.
>>>      
>>>
>
>
>  
>