[gram-user] Segmentation fault from globusrun-ws

[Aragorn Steiger]

This did fix the problem.

Thanks alot!!


Charles Bacon wrote:
> Can you apply the globus_xio-0.38 update from 
> http://www.globus.org/toolkit/advisories.html?  That should fix this 
> problem.
>
>
> Charles
>
> On May 28, 2008, at 12:32 PM, Aragorn Steiger wrote:
>
>>
>> This is in regards to testing globusrun-ws per 2.7 in the quickstart 
>> guide: 
>> http://www.globus.org/toolkit/docs/4.0/admin/docbook/quickstart.html#q-gram 
>>
>> When I run globusrun-ws I get:   Segmentation fault
>> I get this no matter the combination of options.
>>
>> I am installing globus-4.0.7 with jdk1.5.0_09, gcc-3.4.6, 
>> postgres-8.3.0 and ant-1.7.0 on RHEL4 2.6.9-42.ELsmp x86_64 GNU/Linux.
>>
>> Build Process
>> I compiled from source using the steps from the Quickstart guide with 
>> the following modifications.
>> I compiled with these options:
>> ./configure --with-buildopts="-verbose" --enable-wsgram-pbs 
>> --with-iodbc=/wsu/usr/local/lib
>> I ran setup-globus-gram-job-manager after the hostcert.pem was in place.
>>
>> For the container.log file that I have included the two commands 
>> below were issued by user at8036 shortly after globus was restarted. 
>> container.log has output from SOAP Message Logging, GRAM debugging, 
>> and RFT debugging as well as the default options.
>>
>> [at8036 at Commodore ch_p4]$ globusrun-ws -submit -c /bin/true
>> Segmentation fault
>> [at8036 at Commodore ch_p4]$ globusrun-ws
>> Segmentation fault
>>
>> The permissions on all of the host certs are good.
>> Modulus on the host certs is good.
>>
>> I ran setup-globus-gram-job-manager after the hostcert file was 
>> created per the Warning generated during the install. I guessed that 
>> this was to be run with root permission since there is a non-root 
>> option for setup-globus-gram-job-manager. Is it possible that I 
>> should have run this as globus? If so how do I repair this?
>> Warning: Host cert file: /etc/grid-security/hostcert.pem not found.  
>> Re-run
>>      setup-globus-gram-job-manager after installing host cert file.
>>
>> I was able to submit jobs using managed-job-globusrun. Although these 
>> jobs appeared to complete successfully, they did not actually produce 
>> the expected output. Again, I may not be submitting things properly 
>> so I don't know for sure if managed-job-globusrun was not doing what 
>> it was supposed to do. Also, this command is in 
>> ~globus/test/globus_wsrf_gram_service_java_test_unit. Since this is 
>> in a test directory, maybe it is not meant to be used? I just tried 
>> it because I saw someone mentioning it in a post, and found it in a 
>> tutorial.
>>
>> The gdb report for this is as follows:
>> [at8036 at Commodore ~]$ gdb globusrun-ws
>>
>> (gdb) r
>> Starting program: /wsu/home/groups/globus/bin/globusrun-ws
>>
>> Program received signal SIGSEGV, Segmentation fault.
>> 0x0000002a984714d0 in globus_l_xio_udp_write 
>> (driver_specific_handle=Cannot access memory at address 0x7effffffe8
>> ) at globus_xio_udp_driver.c:1294
>> 1294                    result = globus_l_xio_udp_libc_convert_addr(
>>
>> (gdb) n
>>
>> Program terminated with signal SIGSEGV, Segmentation fault.
>> The program no longer exists.
>>
>> Also, the dmesg error for these faults is as follows:
>> globusrun-ws[8535]: segfault at 0000007effffff24 rip 0000002a984714d0 
>> rsp 0000007fbfffdd90 error 6
>>
>> -- 
>> Thank You,
>>
>>    Aragorn Steiger
>>    Wayne State University
>>    Computing & Information Technology
>>    (313) 577-9601