Fix thanks to Chong Chen at IBM - "When invoking the LSF drmma api to submit a job, the API will print the message like 'Job <78> is submitted to default queue <normal>'. So if the Galaxy want to use this API, need modify the drmaa_external_runner.py script to handle this message. LSF already support a env variable to disable this message print out, "BSUB_QUIET", set this into env can disable this message output."
A second issue was related to subsequent querying of the job state as a user other than the submitter - according to Chong Chen this has been fixed in latest LSF DRMAA changes on github (https://github.com/PlatformLSF/lsf-drmaa/commit/43e62546172b90d9f95e90273d77f4c35a0bb404).
To obtain the behavior (assigning all groups for a given user for each job submission) change the property - 'drmaa_external_runjob_script' from 'scripts/drmaa_external_runner.py' to 'scripts/drmaa_external_runner.py --assign_all_groups'.
Making this optional because my concern is that this could be an expensive operation on some clusters.
- The secondary user groups are not assigned to the user
- When the json file is on a NFS share where root access is not allowed this script fails with a "error: JobTemplate file (/path/to/jsonfile) doesn't exist" error
To fix the first we have to go through all the groups of the user in set_user(uid) and assign them with os.setgroups()
The second one is fixed by removing the check if the json file exists in the function validate_paramters and move that check to a new function which is called after set_user(uid). Because its possible that the new user (the one in uid) has access to that file but root hasn't
John Chilton