Changeset 1600 for trunk/src/lib/Components/qsim.py

Timestamp:

07/02/09 13:17:06 (9 months ago)

Author:

wtang

Message:

change 'location' attribute from string to list, corresponding to the change of new cqm

Files:

• trunk/src/lib/Components/qsim.py (modified) (15 diffs)

trunk/src/lib/Components/qsim.py

@@ -146 +146 @@
     end_time: unix second, float
     failure_time: unix second, float
-    location: 'partition:partition', string
+    location: list of string(partition name)
     state: ['invisible', 'running', 'queued', 'ended', 'pending']  string
     is_visible: true/false
@@ -201 +201 @@
         self.recovery_opt = spec.get("recovery_opt", RECOVERYOPT)
         self.checkpoint = 1
-        self.location = ''
+        self.location = []
 
 class JobList(DataList):
@@ -332 +332 @@
         
         #initialize time stamps and job queues
-        #time stamp format: ('EVENT', 'time_stamp_date', time_stamp_second, {'job_id':str(jobid), 'location':'partition-name'})
+        #time stamp format: ('EVENT', 'time_stamp_date', time_stamp_second, {'job_id':str(jobid), 'location':[partition1, partition2,...]})
         self.time_stamps = [('I', '0', 0, {})]
         self.cur_time_index = 0
@@ -536 +536 @@
             message = "%s;Q;%d;queue=%s" % (timestamp, spec['jobid'], spec['queue'])
         elif eventtype == 'R':  #resume running after failure recovery
-            message = "%s;R;%s" % (timestamp, spec['location'])
+            message = "%s;R;%s" % (timestamp, ":".join(spec['location']))
         else:
             wall_time = spec['walltime']
@@ -545 +545 @@
                 message = "%s;S;%d;queue=%s qtime=%s Resource_List.ncpus=%s Resource_List.walltime=%s start=%s exec_host=%s" % \
                 (timestamp, spec['jobid'], spec['queue'], spec['submittime'], 
-                 spec['nodes'], log_walltime, spec['start_time'], spec['location'])
+                 spec['nodes'], log_walltime, spec['start_time'], ":".join(spec['location']))
             elif eventtype == 'E':  #end
                 message = "%s;E;%d;queue=%s qtime=%s Resource_List.ncpus=%s Resource_List.walltime=%s start=%s end=%f exec_host=%s runtime=%s" % \
                 (timestamp, spec['jobid'], spec['queue'], spec['submittime'], spec['nodes'], log_walltime, spec['start_time'], 
-                 round(float(spec['end_time']), 1), spec['location'], 
+                 round(float(spec['end_time']), 1), ":".join(spec['location']),
                  spec['runtime'])
             elif eventtype == 'F':  #failure
@@ -555 +555 @@
                 message = "%s;F;%d;queue=%s qtime=%s Resource_List.ncpus=%s Resource_List.walltime=%s exec_host=%s start=%s frag_runtime=%s complete=%f" % \
                 (timestamp, spec['jobid'], spec['queue'], spec['submittime'], 
-                 spec['nodes'], log_walltime, spec['location'], spec['start_time'], 
+                 spec['nodes'], log_walltime, ":".join(spec['location']), spec['start_time'], 
                  frag_runtime, round(frag_runtime / float(spec['runtime']), 2)
                 )
@@ -561 +561 @@
                 message = "%s;P;%d;queue=%s qtime=%s Resource_List.ncpus=%s Resource_List.walltime=%s exec_host=%s start=%s" % \
                 (timestamp, spec['jobid'], spec['queue'], spec['submittime'], 
-                 spec['nodes'], log_walltime, spec['location'], spec['start_time'], 
+                 spec['nodes'], log_walltime, ":".join(spec['location']), spec['start_time'], 
                 )
                 print "message=", message
@@ -595 +595 @@
             
             #release partition immediately
-            partitions = jobspec['location'].split(':')
+            partitions = jobspec['location']
             for partition in partitions:
                 self.release_partition(partition)
@@ -613 +613 @@
   
             #release partition
-            partitions = jobspec['location'].split(':')
+            partitions = jobspec['location']
             for partition in partitions:
                 print "partition %s start repairing" % (partition)
@@ -625 +625 @@
             failure_datetime = sec_to_date(fail)
             self.log_job_event('F', failure_datetime, jobspec)
-            print self.get_current_time(), " job %d failed at %s!!" % (job_id, jobspec['location'])
+            print self.get_current_time(), " job %d failed at %s!!" % (job_id, ":".join(jobspec['location']))
             
             rec_updates = self.recovery_mgr(jobspec)
@@ -714 +714 @@
         '''update the job state and start_time and end_time when cqadm --run
         is issued to a group of jobs'''
-        partitions = updates['location'].split(':')
+        partitions = updates['location']
         for partition in partitions:
             self.reserve_partition(partition)
@@ -839 +839 @@
         print "run job specs=", specs, " on partion", nodelist
         if specs:
-            self.start_job(specs, {'location': ":".join(nodelist)})
+            self.start_job(specs, {'location': nodelist})
             #set tag false, enable scheduling another job at the same time
             self.increment_tag = False
@@ -867 +867 @@
         return midplane_list 
     
-    def get_next_failure(self, location, now, duration):    #change for rarm
+    def get_next_failure(self, location, now, duration): 
         '''return the next(closest) failure moment according the partition failure list'''
         
@@ -884 +884 @@
                                        
         closest_fail_sec = MAXINT
-        partitions = location.split(':')
-        
+        partitions = location
+
         midplanes = set()
         for partition in partitions:
@@ -1183 +1183 @@
         job_end_times = {}
         for item in end_times:
-            job_end_times[item[0]] = item[1]
-            
+            job_end_times[item[0][0]] = item[1]
+                    
         now = self.get_current_time_sec()
         for p in self.cached_partitions.itervalues():