Re: [HTCondor-users] GAHP fails INITIALIZE_FROM_FILE / how to use separate globus libs

["Fischer, Max (SCC)" <max.fischer@xxxxxxx>]

Hi Jaime,

thanks for the insight and suggestion!

I have ended up adding
	export LD_LIBRARY_PATH='/usr/lib64/condor/'
to /etc/sysconfig/condor, which is sourced by the init script.

For some reason, setting GRIDMANAGER_ENVIRONMENT did *not* work, even though the config is definitely read.
Since the process shuts down quickly, I haven't been able to inspect the GRIDMANAGER to see whether there was a problem in how it was set, or if it was set at all.

Cheers,
Max

> Am 16.12.2016 um 16:15 schrieb Jaime Frey <jfrey@xxxxxxxxxxx>:
> 
> I suspect you have Globus 6.0 library packages installed. HTCondor 8.4.x is linked against Globus 5.2.5, which is incompatible but has the same SO version on the libraries. HTCondor 8.5.3 and above are linked against Globus 6.0, so those versions should work with your system Globus libraries.
> 
> If you canât upgrade to the latest 8.5.x release, you can modify your condor init script to set LD_LIBRARY_PATH in the condor_masterâs environment (which will be inherited by the other daemons).
> Or you can set this in your HTCondor configuration file to get the same effect for the condor_gridmanager and gahps:
>  GRIDMANAGER_ENVIRONMENT = âLD_LIBRARY_PATH=/usr/lib64/condorâ
> 
> - Jaime
> 
>> On Dec 16, 2016, at 8:52 AM, Fischer, Max (SCC) <max.fischer@xxxxxxx> wrote:
>> 
>> Hi Brian,
>> 
>> sure, sorry for not mentioning it:
>> 
>> condor: 8.4.10 from repo htcondor-stable
>> 	- same for condor-all, condor-cream-gahp
>> globus-rsl: 10.11 from repo epel
>> 	- probably the same for other globus, didn't check them all
>> 
>> Cheers,
>> Max
>> 
>>> Am 16.12.2016 um 14:59 schrieb Brian Bockelman <bbockelm@xxxxxxxxxxx>:
>>> 
>>> Hi Max,
>>> 
>>> What version of HTCondor?
>>> 
>>> [I know Jaime has continuously worked to improve the linking (it's pretty nightmarish to combine all these platforms together...); I just want to make sure we don't chase old bugs.]
>>> 
>>> Brian
>>> 
>>>> On Dec 16, 2016, at 6:55 AM, Fischer, Max (SCC) <max.fischer@xxxxxxx> wrote:
>>>> 
>>>> Hi all,
>>>> 
>>>> I'm seeing the following errors in the logs when trying grid submission to cream or nordugrid resources:
>>>> 
>>>> 12/16/16 11:29:39 (pid:3381524) (D_ALWAYS) [3381524] GAHP server pid = 3381561
>>>> 12/16/16 11:29:39 (pid:3381524) (D_ALWAYS) [3381524] GAHP command 'INITIALIZE_FROM_FILE' failed: -1
>>>> 12/16/16 11:29:39 (pid:3381524) (D_ALWAYS) [3381524] GAHP: Failed to initialize from file
>>>> 
>>>> From what I've found in the GAHP docs, it's expecting my proxy there. Doing this manually fails with the same error code and a message related to Globus:
>>>> 
>>>> $ nordugrid_gahp
>>>> $GahpVersion: 1.3.0 Dec 13 2016 Nordugrid\ Gahp $
>>>> INITIALIZE_FROM_FILE /tmp/x509up_u12080
>>>> F -1 Failed\ to\ activate\ Globus modules
>>>> 
>>>> Digging through HTCondor-users, I've found a thread [1] attributing such failures to incorrect resolution of Globus libraries. The tread mentions similar errors when the system globus is used instead of condor's own.
>>>> Andrew Lahiff asked for a status on this at the beginning of the year [2], but it's been several versions since then.
>>>> 
>>>> Anyways, using ldd indeed gives for example
>>>> 	libglobus_common.so.0 => /usr/lib64/libglobus_common.so.0
>>>> instead of
>>>> 	libglobus_common.so.0 => /usr/lib64/condor/libglobus_common.so.0
>>>> 
>>>> Adding a /etc/ld.so.conf.d/ and ldconfig as in [1] still prefers the system globus libraries.
>>>> Setting the dependency explicitly via
>>>> 	export LD_LIBRARY_PATH='/usr/lib64/condor/'
>>>> shows the correct libraries with ldd and manually using gahp, INITIALIZE_FROM_FILE succeeds.
>>>> 
>>>> Is there a suggested way to have only condor use its own globus libraries? We would really like to keep the separate globus tools.
>>>> Can this be set in condor's configuration, without changing it for the rest of the system?
>>>> 
>>>> Cheers,
>>>> Max
>>>> 
>>>> [1] [HTCondor-users] CREAM error: Failed to start gahp
>>>> https://lists.cs.wisc.edu/archive/htcondor-users/2015-September/msg00072.shtml
>>>> 
>>>> [2] Re: [HTCondor-users] CREAM error: Failed to start gahp
>>>> https://lists.cs.wisc.edu/archive/htcondor-users/2016-January/msg00108.shtml_______________________________________________
>>>> HTCondor-users mailing list
>>>> To unsubscribe, send a message to htcondor-users-request@xxxxxxxxxxx with a
>>>> subject: Unsubscribe
>>>> You can also unsubscribe by visiting
>>>> https://lists.cs.wisc.edu/mailman/listinfo/htcondor-users
>>>> 
>>>> The archives can be found at:
>>>> https://lists.cs.wisc.edu/archive/htcondor-users/
>>> 
>>> _______________________________________________
>>> HTCondor-users mailing list
>>> To unsubscribe, send a message to htcondor-users-request@xxxxxxxxxxx with a
>>> subject: Unsubscribe
>>> You can also unsubscribe by visiting
>>> https://lists.cs.wisc.edu/mailman/listinfo/htcondor-users
>>> 
>>> The archives can be found at:
>>> https://lists.cs.wisc.edu/archive/htcondor-users/
>> 
>> _______________________________________________
>> HTCondor-users mailing list
>> To unsubscribe, send a message to htcondor-users-request@xxxxxxxxxxx with a
>> subject: Unsubscribe
>> You can also unsubscribe by visiting
>> https://lists.cs.wisc.edu/mailman/listinfo/htcondor-users
>> 
>> The archives can be found at:
>> https://lists.cs.wisc.edu/archive/htcondor-users/
> 
> 
> _______________________________________________
> HTCondor-users mailing list
> To unsubscribe, send a message to htcondor-users-request@xxxxxxxxxxx with a
> subject: Unsubscribe
> You can also unsubscribe by visiting
> https://lists.cs.wisc.edu/mailman/listinfo/htcondor-users
> 
> The archives can be found at:
> https://lists.cs.wisc.edu/archive/htcondor-users/

Attachment: smime.p7s
Description: S/MIME cryptographic signature