[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [HTCondor-users] Local universe in DAGs

> I'm using condor 8.3.7 with 2 worker nodes and a 'personal' (i.e. everything) host all on Ubuntu 14.04. 

> In submitting DAG jobs, I have noticed that if one node uses the 'local' universe, then the job completes successfully (according to the Schedd log), but the dag.nodes.log file is not written to. 

> This means that any child nodes of this node will never be kicked off and the dag sits around waiting forever. 

> Is this expected behaviour? If so, then why? 

No, it's not!!

> Note: I have tried this on our production setup, which uses Condor 8.2.4, and it works as expected, so I assume this is a bug after this version, as opposed to new functionality? The problem is also in the git master (8.5.1 at current). 

Yes, this is a bug.

It turns out that there is a workaround, though:  if you define a log file (it doesn't matter what) in your local universe submit file, the nodes.log file will get correctly written to.  (This is actually why we didn't see the problem, because we have a test that tests local universe with DAGMan, but the submit file defines a log file.)  It even works if your log file is /dev/null!

I'll generate a bug ticket for this, and we'll take a look at it ASAP.

Kent Wenger