Details
-
Task
-
Status: Open
-
Minor
-
Resolution: Unresolved
-
None
-
None
-
None
-
None
Description
For near real-time jobs running on hadoop, we want to minimize impact on them when rolling restarting tasktrackers. The idea here is to restart tasktrackers from jobtracker and selectively choose task trackers according to tasks' status on them to do rolling restart, such that the impact on total job running time is minimum.