[
https://issues.apache.org/jira/browse/MAPREDUCE-1001?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=12906039#action_12906039
]
Greg Roelofs commented on MAPREDUCE-1001:
-
Hong's maintenance concerns have been borne out; here's my comment from
(duplicate) MAPREDUCE-2024:
bq. The fifth JobTracker constructor (JobTracker(final JobConf conf, Clock
clock, boolean ignoredForSimulation)) still uses the old mapred.* config
settings and appears to duplicate much of the code in the fourth (main) ctor.
It should be modernized and, ideally, share as much code as possible with the
main one in order to minimize simulation drift and the potential for config
errors.
(In case it's not clear, the other constructor has already been modified to use
the newer mapreduce.* config settings. Thus the config settings that work
for one will have no effect on the other unless they're manually duplicated.)
Reducing code duplication in Mumak
--
Key: MAPREDUCE-1001
URL: https://issues.apache.org/jira/browse/MAPREDUCE-1001
Project: Hadoop Map/Reduce
Issue Type: Bug
Reporter: Hong Tang
The first release of Mumak (MAPREDUCE-728) introduces some coupling between
the core mapred code with Mumak code. Specifically, New constructors are
added to JobTracker and JobInProgress to allow simulator to subclass and
alter JT/JIP behavior. This could be a code maintenance overhead when new
changes have to be ported to either the added constructors or the simulation
subclasses.
It would be nice to refactor the constructors of JobTracker and JobInProgress
to avoid as much code duplication as possible.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
