Is there any way to stop or kill a 'Stranded' Glide docking run but still recover any docking results and scores that were completed?

When a job is 'Stranded', it means that the Job Control process responsible for the computational backend died (or can't be communicated with), so Job Control is unable to determine the status of the backend. This could be caused by a machine crashing or rebooting, but in other cases, the backend (i.e., the Glide docking process) might still be running.

Eventually,the Job Control system will attempt recovery automatically, at which point the job status should changed to 'died' if the job failed. If you're running a distributed Glide job and it's a subjob, not the main driver, that is 'stranded', the driver should rerun the failed subjob automatically.

You can hasten the recovery process by running

$SCHRODINGER/jobcontrol -recover JobId

where JobId is the Schrodinger Job ID, visible in the Monitor panel or near the top of the log file. If this is a serial job, before you run this command, try looking at the scratch directory for the job (tmpdir/username/jobname/), as there should be a jobname_raw.maegz file there that contains the docking results so far.

There is a problem in Glide 5.5 that prevents this intermediate results file from being transferred to the launch directory when a job fails, so please copy this file to a safe location before attempting recovery. This issue is resolved in Glide 5.6 (Suite 2010).

With the intermediate 'raw' file, you then could run a new Glide job on the ligands that hadn't yet docked (e.g., by copying the jobname.in file and using the LIGAND_START keyword to indicate the starting ligand number), and then combine the results with those of the original job. To combine the results, first run $SCHRODINGER/utilities/glide_sort on the 'raw' file to sort it, and then run $SCHRODINGER/utilities/glide_merge on the old and new 'pv' or 'lib' files to generate the final pose file.

Keywords: recover Glide results, stranded docking run, job control, pause, restart

Back to Search Results