 |
 |
On Tue, 29 May 2007, Stijn De Smet wrote: > On my UI, the jobs remain in the scheduled state, while they are > running(in the logging-info the Running is there) The Running and Done events from the WN are sent almost directly to the LB Server running on the RB, but do not count in deciding what the current state of the job is (in the past they did, but that led to confusion in the state machine, and there was no strong reason to complicate it further). As soon as the job has been submitted to Condor-G on the RB, the state of the job is decided by the Log Monitor process, which is monitoring the logs of Condor-G, which itself gets updates from grid_monitor processes running on CEs. If the RB is busy with many jobs, it can be a while before Condor-G gets to the Running or Done event for any particular job, which may have finished long ago. I suppose you do get the events in the end? This (slightly simplified) diagram can be of service here: http://goc.grid.sinica.edu.tw/gocwiki/Job_submission_chain_diagram