[Swift-devel] Problems running coaster

[Michael Wilde]

I tried NCSA-HG instead (to avoid the issues of the UC two-architecture 
site).

The coaster provider works for me there, using gt2:gt2:pbs, as you 
reported previously.  Same Swift script that fails on abe.

So it looks like something in the job specs that is launching coaster 
for gt2:pbs is not being accepted by abe.

I also see that these log messages which I mentioned earlier do not 
occur on the successful mercury coaster run:

-- 
2008-07-28 08:22:58,853-0500 INFO  ServiceManager Service task 
Task(type=JOB_SUBMISSION, identity=urn:cog-1217251364678) terminated. 
Removing service.
2008-07-28 08:22:58,853-0500 INFO  ServiceManager Service does not 
appear to be registered with this manager
-- 

If someone can point me closer to where the swift job that launches the 
bootstrap script is run from, and how, I can try to reproduce the 
problem with globus-job-run or globus-run.  In the meantime I will hunt 
for that.

- Mike


On 7/28/08 11:52 AM, Michael Wilde wrote:
> 
> 
> On 7/28/08 9:31 AM, Ben Clifford wrote:
>> On Sun, 27 Jul 2008, Michael Wilde wrote:
>>
>>> Yes. Just *before* I tried abe with coasters, I did a simple 
>>> globus-job-run to
>>> its pbs jobmanager. That worked fine.
>>
>> try submitting with the following changes to sites.xml:
>>
>>  1. not using coasters, instead using the gt2 provider
>>  2. not specifying tg allocation in sites.xml
>>
> 
> Using globus-job-run I see that only 1 account is valid.
> 
> globus-job-run works to jobmanager-pbs, both with and without a -p 
> option specifying the valid account.
> 
> Swift using the gt2 provider also works with and without the valid 
> account specified in sites.xml as a globus profile.
> 
> Swift and the coaster provider with gt2:gt2:pbs fails both without the 
> valid account (it gets the same "This node is in dedicated user mode" 
> message)
> 
> In terms of sites.xml:
> 
> This fails (using coaster provider):
> 
> <config>
> <pool handle="abe" >
>   <execution provider="coaster" url="grid-abe.ncsa.teragrid.org" 
> jobManager="gt2:gt2:pbs" />
>   <gridftp url="gsiftp://gridftp-abe.ncsa.teragrid.org"/>
>   <workdirectory>/u/ac/wilde/swiftwork</workdirectory>
> </pool>
> </config>
> 
> This works (using gt2 provider):
> 
> <config>
> <pool handle="abe">
>   <jobmanager universe="vanilla" 
> url="grid-abe.ncsa.teragrid.org/jobmanager-pbs" major="2"/>
>   <gridftp url="gsiftp://gridftp-abe.ncsa.teragrid.org"/>
>   <workdirectory>/u/ac/wilde/swiftwork</workdirectory>
> </pool>
> </config>
> 
> I will try the same to a different TG site (starting with UC).
> 
> - Mike
> 
>