[mpich-discuss] assertion failed in socksm.c it_plfc->revents when trying mpptest

[Gregory Alan Hildstrom]

Aha! No it does not. It errors with the same error, so it is not an 
mpptest problem.

Process 0 of 2 is on servernode
Process 1 of 2 is on computenode
Assertion failed in file socksm.c at line 1663: (it_plfd->revents & 
0x008) == 0
internal ABORT - process 0
rank 0 in job 2  servernode_54275   caused collective abort of all ranks
   exit status of rank 0: return code 1

I discovered that my /etc/hosts file was incorrect on the computenode. I 
changed it to match the servernode's /etc/hosts file and things seem to 
be working correctly now. I was able to run cpi and mpptest.

Thank you. -Greg

Pavan Balaji wrote:
> 
> The problem with the test suite seems quite weird to me. We run the 
> MPICH2 test suite several times in the nightly scripts as well as for 
> our local testing every day, and we never saw any such problem.
> 
> Does the "cpi" test in your build/examples directory work fine?
> 
>  $ mpiexec -n 4 ./examples/cpi
> 
>  -- Pavan
> 
> On 06/10/2009 06:07 PM, Gregory Alan Hildstrom wrote:
>> Hello. I am trying to set up a simple performance test using mpptest 
>> between two nodes running a stripped down version of RHEL5 x86_64. I 
>> am able to mpdboot -n 2 and the subsequent mpdtrace shows the host 
>> names of both nodes. I ran mpdringtest with no problems. I'm running 
>> mpich2-1.1 and mpptest from perftest-1.4b.
>>
>> I ran "mpirun -np 2 mpptest -logscale" and got the following error 
>> output.
>>
>> Assertion failed in file socksm.c at line 1663: (it_plfd->revents & 
>> 0x008) == 0
>> internal ABORT - process 0
>> Assertion failed in file socksm.c at line 1663: (it_plfd->revents & 
>> 0x008) == 0
>> internal ABORT - process 1
>> rank 0 in job 3  serverhost_37494   caused collective abort of all ranks
>>   exit status of rank 0: return code 1
>>
>> I am not sure if this is an mpich issue or an mpptest issue. I tried 
>> to make the test suite bundled with mpich2, but it did not build, 
>> which is why I am trying mpptest. The test suite first complained 
>> about a missing lib.a, which should have been libmpich.a in 
>> test/mpid/ch3's makefile. Then it complained about a missing test.h in 
>> test/util/timer/timers.c. I commented out the include test.h, which 
>> resulted in an undefined reference to "Test_Waitforall".
>>
>> Any suggestions of where to look for problems? I would also appreciate 
>> any pointers to other mpich-compatible benchmarks or test suites.
>>
>> Thank you. -Greg
>>
> 

-- 
Gregory Alan Hildstrom
Senior Secure Systems Engineer
Trusted Computer Solutions
10010 San Pedro Ave, Suite 220
San Antonio, TX 78216
Office: 210-340-3151x101
Mobile: 210-413-6082
Fax: 210-340-3568
Email: ghildstrom at trustedcs.com
Email2: hildstrom at yahoo.com