Move things around a little in the main loop.

Should take care of two things:

1. When looking at the DEBUG log level, the status should appear before we
   wait for WQ, giving the user time to digest it, and not have to look for
   it, scrolling through tons of task creation output.
2. Plotting should also be triggered before WQ is polled for tasks, which
   minimizes the potential database access overlap.

lobster/commands/process.py

@@ -241,14 +241,6 @@ def sprint(self):
                 logger.info("terminating gracefully")
                 break
 
-            stats = self.queue.stats_hierarchy
-            logger.info("{0} out of {1} workers busy; {3} tasks running, {4} waiting; {2} units left".format(
-                    stats.workers_busy,
-                    stats.workers_busy + stats.workers_ready,
-                    units_left,
-                    stats.tasks_running,
-                    stats.tasks_waiting))
-
             expiry = None
             if proxy:
                 left = proxy.getTimeLeft()
@@ -266,6 +258,7 @@ def sprint(self):
                 have[c] = cstats.tasks_running + cstats.tasks_waiting
 
             t = time.time()
+            stats = self.queue.stats_hierarchy
             tasks = task_src.obtain(stats.total_cores, have)
 
             for category, cmd, id, inputs, outputs, env, dir in tasks:
@@ -297,7 +290,22 @@ def sprint(self):
                 self.queue.submit(task)
             creation_time += int((time.time() - t) * 1e6)
 
+            stats = self.queue.stats_hierarchy
+            logger.info("{0} out of {1} workers busy; {3} tasks running, {4} waiting; {2} units left".format(
+                    stats.workers_busy,
+                    stats.workers_busy + stats.workers_ready,
+                    units_left,
+                    stats.tasks_running,
+                    stats.tasks_waiting))
+
             task_src.update(self.queue)
+
+            # recurring actions are triggered here; plotting etc should run
+            # while we have WQ hand us back tasks w/o any database
+            # interaction
+            if action:
+                action.take()
+
             starttime = time.time()
             task = self.queue.wait(interval)
             tasks = []
@@ -329,10 +337,6 @@ def sprint(self):
                 logger.info("activating fast abort with multiplier: {0}".format(abort_multiplier))
                 abort_active = True
                 self.queue.activate_fast_abort(abort_multiplier)
-
-            # recurring actions are triggered here
-            if action:
-                action.take()
         if units_left == 0:
             logger.info("no more work left to do")
             if action: