[MAPREDUCE-4001] Improve MAPREDUCE-3789's fix logic by looking at job's slot demands instead

• Type: Improvement
• Status: Closed
• Priority: Minor
• Resolution: Fixed
• Affects Version/s: 1.1.0
• Fix Version/s: 1.1.0
• Component/s: capacity-sched
• Hadoop Flags: Reviewed
• Assignee: Harsh J
• Reporter: Harsh J
• Created: 13/Mar/12 13:11
• Updated: 17/Oct/12 18:27
• Resolved: 13/Mar/12 21:32

Description

In MAPREDUCE-3789, the fix had unfortunately only covered the first time assignment scenario, and the test had not really caught the mistake of using the condition of looking at available TT slots (instead of looking for how many slots a job's task demands).

We should change the condition of reservation in such a manner:

          if ((getPendingTasks(j) != 0 &&
               !hasSufficientReservedTaskTrackers(j)) &&
-                (taskTracker.getAvailableSlots(type) !=
+                !(j.getNumSlotsPerTask(type) >
                  getTTMaxSlotsForType(taskTrackerStatus, type))) {

I had not realized during the earlier ticket that j.getNumSlotsPerTask(type) did exist.

Issue Links

• relates to: MAPREDUCE-516 Fix the 'cluster drain' problem in the Capacity Scheduler wrt High RAM Jobs