[Mckernel-users 34] Re: [Problem with installation of mckernel again]

[안민우]

To be specific, Linpack benchmark problem shows

kernel:BUG: soft lockup - CPU#0 stuck for 23s! [mcexec:63572]
Terminate by signal 28

2017-05-26 12:52 GMT+09:00 안민우 <mwahn402 at gmail.com>:

> Hello Balazs,
>
> I solved all problems I send before.
>
> I compared elapsed time of each workloads in NPB-MPI, results was quite
> strange.
>
> My Intel Xeon server have 2 NUMAs and 6 cores each. I assigned 6 cores
> from NUMA1 and 2 cores from NUMA0(total 8 cores) to McKernel and run NPB.
> Compared to this, I used same cores(8 cores) to run NPB in CentOS.
>
> According to my result, elapsed time of NPB on McKernel vs. Linux was
> similar, but every workloads show that elapsed time of NPB on Linux is
> little bit shorter. Is it right result?
>
>
> Moreover, I have some runtime error to run "Linpack" benchmark. Linpack
> stops it's running during runtime.
>
> I use mpiexec command as $mcexec mpiexec -np 8 <executable>.
>
> Is there any problem? And there is way to see kmsg on McKernel before my
> server reboots itself?
>
>
> Thank you to read my email.
>
> Best regards,
>
> Minwoo Ahn.
>
> 2017-05-17 11:18 GMT+09:00 안민우 <mwahn402 at gmail.com>:
>
>> Hello Balazs,
>>
>> I'm MS student from Korea, who sent emails about installation of
>> mckernel(I successfully installed before). I have 3 questions.
>>
>> 1.
>>
>> I have some problem in my linux, so I install centOS again and I tried to
>> build and boot mckernel.
>>
>> I successed to build mckernel(ihk+mckernel.tgz as you gave to me), but
>> when I tried to run mcreboot.sh script in "install/sbin" directory, my
>> whole server rebooted itself.
>>
>> I'm sure that I disabled irqbalance and changed to SELINUX=disabled.
>>
>> How can I fix my problem?
>>
>> 2.
>>
>> Moreover, how can I find some documents(or manual) about mckernel source
>> code?
>>
>> It is for change # of cores of mckernel. In my server, without any
>> modification of configuration, mckernel installed on 6 cores. Can I change
>> it to 8 cores?
>>
>> 3.
>>
>> I tried to run NAS parallel benchmark in MPI version(ft.B.16), it causes
>> page fault(What's the problem?). Is it write to run "ft.B.16" executable
>> file with command "MPI_NUM_THREAD=16 mcexec {PATH_TO_EXECUTABLE}/ft.B.16" ?
>>
>>
>>
>> Thank you to read my email.
>>
>> Best regards,
>>
>> Minwoo Ahn.
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.pccluster.org/pipermail/mckernel-users/attachments/20170526/f186c244/attachment.html>