hadoop-yarn-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Sangjin Lee <sj...@apache.org>
Subject [YARN-2928] rebasing to trunk
Date Tue, 13 Oct 2015 01:02:55 GMT
Hi folks,

I have completed the rebase of YARN-2928 (this time cherry-picks really) to
the trunk as of last Saturday. I resolved 10 merge conflicts most of which
were minor. But I do want to call out a few of them, and would like you to
review how I resolved those conflicts before I make the rebase official. I
have just pushed this new branch ("*YARN-2928-rebase*") so you can take a
look at it. I'll swap the branches once we're satisfied.

The following are those commits to review. I called out those who might be
best to review the merges.

[3e3a8fe: Junping]
Trunk added a new use (in TestContainerResourceIncreaseRPC
of a method (TestRPC.newContainerToken) in TestRPC which we moved from
yarn-common to yarn-server-common. I copied that method in
TestContainerResourceIncreaseRPC. We could reconsider whether we want to
move TestRPC from yarn-common to yarn-server-common. I don't recall the
details of the discussion, but was there a strong reason to move TestRPC
out of yarn-common? If trunk keeps creating new uses of this class, it
might be a problem.

[d35d861: Naga]
Trunk added a new RM event type (app updated: YARN-4044
I applied the same changes and moved code to
AbstractTimelineServicePublisher, TimelineServiceV1Publisher, and
TimelineServiceV2Publisher respectively. Naga, could you please confirm if
that new event is done right in the merge commit?

It turns out HDFS-9080 broke the HBase mini-cluster, which in turn broke
our HBase-based unit tests. This was caught by HDFS-9187 which has a patch.
The patch is not entirely correct (causes NPEs), and I applied a fixed
version of that patch to our branch to ensure our tests pass. Let me know
if you are OK with that. I don't think we can wait until HDFS-9187 gets

If you could take a look at these commits, and let me know +1/-1, I'll be
able to take the next steps. Thanks everyone!


  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message