Added '--name' sub-selection to jman list action; some small bug fixes in status of jobs.

gridtk/local.py

@@ -167,7 +167,7 @@ class JobManagerLocal(JobManager):
       while True:
         # Flag that might be set in some rare cases, and that prevents the scheduler to die
         repeat_execution = False
-        # FIRST, try if there are finished processes; this does not need a lock
+        # FIRST, try if there are finished processes
         for task_index in range(len(running_tasks)-1, -1, -1):
           task = running_tasks[task_index]
           process = task[0]
@@ -253,4 +253,4 @@ class JobManagerLocal(JobManager):
         task[0].kill()
         self.stop_job(task[1])
       # stop all jobs that are currently running or queued
-      self.stop_jobs()
+      self.stop_jobs(job_ids)

gridtk/manager.py

@@ -84,7 +84,7 @@ class JobManager:
     q = self.session.query(Job)
     if job_ids:
       q = q.filter(Job.id.in_(job_ids))
-    return list(q)
+    return sorted(list(q), key=lambda job: job.unique)
 
 
   def _job_and_array(self, job_id, array_id = None):
@@ -170,7 +170,7 @@ class JobManager:
     self.unlock()
 
 
-  def list(self, job_ids, print_array_jobs = False, print_dependencies = False, long = False, status=Status, ids_only=False):
+  def list(self, job_ids, print_array_jobs = False, print_dependencies = False, long = False, status=Status, names=None, ids_only=False):
     """Lists the jobs currently added to the database."""
     # configuration for jobs
     if print_dependencies:
@@ -204,7 +204,7 @@ class JobManager:
     self.lock()
     for job in self.get_jobs(job_ids):
       job.refresh()
-      if job.status in status:
+      if job.status in status and names is None or job.name in names:
         print(job.format(format, dependency_length, None if long else 43))
         if print_array_jobs and job.array:
           print(array_delimiter)

gridtk/script/jman.py

@@ -160,7 +160,7 @@ def run_scheduler(args):
 def list(args):
   """Lists the jobs in the given database."""
   jm = setup(args)
-  jm.list(job_ids=get_ids(args.job_ids), print_array_jobs=args.print_array_jobs, print_dependencies=args.print_dependencies, status=args.status, long=args.verbose > 1 or args.long, ids_only=args.ids_only)
+  jm.list(job_ids=get_ids(args.job_ids), print_array_jobs=args.print_array_jobs, print_dependencies=args.print_dependencies, status=args.status, long=args.verbose > 1 or args.long, ids_only=args.ids_only, names=args.names)
 
 
 def communicate(args):
@@ -297,6 +297,7 @@ def main(command_line_options = None):
   # subcommand 'list'
   list_parser = cmdparser.add_parser('list', aliases=['ls'], formatter_class=formatter,  help='Lists jobs stored in the database. Use the -vv option to get a long listing.')
   list_parser.add_argument('-j', '--job-ids', metavar='ID', nargs='+', help='List only the jobs with the given ids (by default, all jobs are listed)')
+  list_parser.add_argument('-n', '--names', metavar='NAME', nargs='+', help='List only the jobs with the given names (by default, all jobs are listed)')
   list_parser.add_argument('-a', '--print-array-jobs', action='store_true', help='Also list the array ids.')
   list_parser.add_argument('-l', '--long', action='store_true', help='Prints additional information about the submitted job.')
   list_parser.add_argument('-x', '--print-dependencies', action='store_true', help='Print the dependencies of the jobs as well.')

gridtk/sge.py

@@ -109,12 +109,17 @@ class JobManagerSGE(JobManager):
     jobs = self.get_jobs(job_ids)
     for job in jobs:
       job.refresh()
-      if job.status in ('queued', 'executing'):
+      if job.status in ('queued', 'executing', 'waiting'):
         status = qstat(job.id, context=self.context)
         if len(status) == 0:
           job.status = 'failure'
           job.result = 70 # ASCII: 'F'
           logger.warn("The job '%s' was not executed successfully (maybe a time-out happened). Please check the log files." % job)
+          for array_job in job.array:
+            if array_job.status in ('queued', 'executing'):
+              array_job.status = 'failure'
+              array_job.result = 70 # ASCII: 'F'
+
 
     self.session.commit()
     self.unlock()