dasahcc commented on a change in pull request #994:
URL: https://github.com/apache/helix/pull/994#discussion_r428323885
##########
File path:
helix-core/src/main/java/org/apache/helix/task/AbstractTaskDispatcher.java
##########
@@ -242,16 +230,16 @@ public void updatePreviousAssignedTasksStatus(
}
break;
case COMPLETED: {
- // The task has completed on this partition. Mark as such in the
context object.
- donePartitions.add(pId);
+ // The task has completed on this partition. Drop it from the
instance and add it to assignedPartitions in
+ // order to avoid scheduling it again in this pipeline.
Review comment:
Also, I dont think you will get DROPPED state as current state, right?
Because the current state will be deleted from the current state if the to
state is DROPPED after it completes state transition.
##########
File path:
helix-core/src/main/java/org/apache/helix/task/AbstractTaskDispatcher.java
##########
@@ -242,16 +230,16 @@ public void updatePreviousAssignedTasksStatus(
}
break;
case COMPLETED: {
- // The task has completed on this partition. Mark as such in the
context object.
- donePartitions.add(pId);
+ // The task has completed on this partition. Drop it from the
instance and add it to assignedPartitions in
+ // order to avoid scheduling it again in this pipeline.
Review comment:
The concern for doing this is that will this reflect the change to
JobContext? So if final state will all be DROPPED. It is hard for us to
determine whether the tasks is completed or it requires another retry? Because
if task has problem in the middle of running, it will be DROPPED as well.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]