I synced up with Junping offline and he found out a couple of issues in the merge. Fixed them along with a couple of other things that I myself found.
I am done with the merge now, just swapped the branches. YARN-2928 is updated and ready now. Created an YARN-2928-old with the old code. Going forward, please continue commits to the (freshly recreated) YARN-2928 branch and I’ll periodically rebase it to trunk. The rebase-to-trunk means that, after every rebase, “git pull —rebase” will get really wierd if you work on the same local branch. So I suggest - Before you pull in the swapped branches into your local environment, take backup of your existing branches. - Use a separate local branch for developing outside of a mirror of origin/YARN-2928 and then cherry-pick right before commit + push. I’ll file tickets for other existing issues in the branch. Thanks +Vinod On Aug 14, 2015, at 8:12 AM, Junping Du <[email protected]<mailto:[email protected]>> wrote: Thanks Vinod for the outstanding effort and Naga, Li for the verification work. I tried to do some diff work between old YARN-2928 branch and YARN-2928-new. I noticed that two branches are based on different commits, so in my local YARN-2928 branch, I manually cherry pick commits between YARN-3543 and HADOOP-12244 (a short list) to make sure two branches are standing at the same points. >From the diff files generated between two branches, I found that most >differences are trivial (blank lines, sequence of methods, etc.). However, >some test failure like TestRMAppLogAggregationStatus is due to our merge >conflict (with YARN-3999) that should be easily fix. I am still working on checking diffs on hints for more test failures, but would like to share with everyone on this branch to watch together in case I miss something. If everyone agree, I would like to create a JIRA to put on the raw diff patch which is open for discussion, and our goal here is to pick up some important pieces of code (not include trivial things) to check-in to YARN-2928-new. Thoughts? BTW, I think TestSystemMetricsPublisherForV2 get NPE failures on both YARN-2928 and YARN-2928-new which should be fixed separately? Thanks, Junping ________________________________ From: Naganarasimha G R (Naga) <[email protected]<mailto:[email protected]>> Sent: Friday, August 14, 2015 3:59 PM To: Li Lu; Vinod Kumar Vavilapalli Cc: [email protected]<mailto:[email protected]>; Zhijie Shen; Junping Du; Sangjin Lee; Vrushali Channapattan Subject: RE: YARN-2928 branch reset Hi Vinod & Lilu, I checked out the new branch compiled and ran Distributed shell everything seems to work fine. Also ran TestDistributedShell , seems to be running fine and also have the fix ready for "TestSystemMetricsPublisherForV2" failure. Shall i raise and apply the patch or you have more issues which you want to raise together with it ? +Naga ________________________________ From: Li Lu [[email protected]<mailto:[email protected]>] Sent: Friday, August 14, 2015 03:32 To: Vinod Kumar Vavilapalli Cc: [email protected]<mailto:[email protected]>; Zhijie Shen; Junping Du; Sangjin Lee; Vrushali Channapattan; Naganarasimha G R (Naga) Subject: Re: YARN-2928 branch reset Thanks for the work Vinod! I’ve tried the latest YARN-2928-new branch with a manually set up cluster and MapReduce jobs. So far it looks good to me. Seems like we’ve quickly recovered to where we stopped on the old YARN-2928 branch (but nothing magically new, of course:) ). I’d propose that everybody working on the YARN-2928 branch to try this out. In this way if there are any problems we can discover that soon. Thanks! Li Lu On Aug 13, 2015, at 14:21, Vinod Kumar Vavilapalli <[email protected]<mailto:[email protected]>> wrote: I had quite a bit of trouble with some of the DistributedShell changes, but finally done with this. Created a new branch YARN-2928-new. If you can give it a spin and things look okay, I’ll swap the branches. Ran TestDistributedShell fine. There are existing failures with tests like TestSystemMetricsPublisherForV2 on the current branch itself, will file tickets for their fixes post the branch-swap. Thanks +Vinod On Aug 12, 2015, at 12:54 PM, Vinod Kumar Vavilapalli <[email protected]<mailto:[email protected]>> wrote: Hey all, I’ve tried rebasing YARN-2928 upto trunk and realized that we haven’t been rebasing correctly. Today, I plan to recreate YARN-2928 branch freshly from trunk. So please hold off any commits to the branch till I send an all-clear. Thanks +Vinod
