Closed rob-c closed 4 years ago
I couldn't agree more! In fact, I believe I have mentioned this a few times in the past. The out of the box behavior is not very Dirac friendly. As I recall you have to have a config with the RuntimePath set to dirac and the Dirac section already present and pointing to the Dirac bashrc and then you also need to set the defaults_DiracProxy->group attribute.
@alexanderrichards I'll check the current docs for how to get a VO certificate but if the current route is for everyone to go via getting a gridpp_user
account then we should strongly consider having all of the defaults in place to be able to submit a hello world
job with these credentials. I'll discuss this with a few people off list here.
I'll get in touch with the LSST users and see what their current configuration is and we should consider adding approved configs to cvmfs as reference (after some thought).
For now I'll drop back to using DIRAC natively but it would be nice if this just worked so I'll try and open some dialogue with LSST people on improving the gridpp docs.
(After a few minutes of poking at this I'm seeing some strange behaviour compared to how I was expecting some things to work so I'd encourage someone checking the latest dev branch for gridpp dirac support in 6.7.0 before it's released).
I'm also seeing:
Ganga In [2]: j=Job();j.backend=Dirac();j.submit()
INFO submitting job 8
INFO job 8 status changed to "submitting"
INFO Preparing Executable application.
INFO Created shared directory: conf-f3f53cb0-6866-4d89-b62b-13f7fd55f681
ERROR ConfigError: attempt to create a configuration section [defaults_GaudiXMLSummary] after bootstrap
ERROR ConfigError: attempt to create a configuration section [defaults_GaudiXMLSummary] after bootstrap ... reverting job 8 to the new status
(Submitting to the local backend works as expected and job completes so will pass tests).
This is from a clean gridpp DIRAC instance with no LHCb ganga environment variables set so I suspect there could be a broken dependency somewhere within the share directory code which is importing Gaudi/LHCb incorrectly on vanilla DIRAC. (This could be my environment being polluted but I'm fairly sure it's not in this case)
That's a new one on me. But then I haven't tried it in a few weeks and there has been a lot go in since then so I'm afraid I can't say if it's just your environment or not. If I get a chance I'll try again today. Either way I think that it proves the point that things aren't behaving as seamlessly as they should.
@alexanderrichards No major rush on the last point it could be some left over thing somewhere in a config file I've just forgotten about, if submitting a gridpp Dirac job works on dev this is just an out of the box config issue
I was hoping to use ganga to test file storage here at ECDF using the standard gridpp tools. (plan was to test the
DiracFile.put
to Edinburgh to test if the xrootd storage protocol is working/supported)Unfortunately I'm seeing the following problems:
From:
/cvmfs/ganga.cern.ch/runGanga-dirac.sh
:Also this fails to launch out of the box when .gangarc doesn't exist due to:
I'll check with lsst to see what they've done in the past but this is not ideal as they now probably have some undocumented solution and in principle on gridpp we don't have an out of the box working
hello world
style job.I think I expect the following...
/cvmfs/ganga.cern.ch/runGanga.sh