Thanks for the new RC, Junping. I built from source and tried it out on
a 2-node cluster with HA enabled. I ran a pi job and some streaming
jobs. I tested that localization and failover work correctly, and I
played a little with the YARN and HDFS web UIs.
I did encounter an old friend of mine, which is that if you submit a
streaming job with input that is only 1 block, you will nonetheless get
2 mappers that both process the same split. What's new this time is that
the second mapper was consistently failing on certain input sizes. I
(re)verified that the issue also exists is 2.7.3, so it's not a
regression. I'm pretty sure it's been there since at least 2.6.0. I
filed MAPREDUCE-6864 for it.
Given that my issue was not a regression, I'm +1 on the RC.
Daniel
On 3/17/17 2:18 AM, Junping Du wrote:
Hi all,
With fix of HDFS-11431 get in, I've created a new release candidate (RC3)
for Apache Hadoop 2.8.0.
This is the next minor release to follow up 2.7.0 which has been released
for more than 1 year. It comprises 2,900+ fixes, improvements, and new
features. Most of these commits are released for the first time in branch-2.
More information about the 2.8.0 release plan can be found here:
https://cwiki.apache.org/confluence/display/HADOOP/Hadoop+2.8+Release
New RC is available at:
http://home.apache.org/~junping_du/hadoop-2.8.0-RC3
The RC tag in git is: release-2.8.0-RC3, and the latest commit id is:
91f2b7a13d1e97be65db92ddabc627cc29ac0009
The maven artifacts are available via repository.apache.org at:
https://repository.apache.org/content/repositories/orgapachehadoop-1057
Please try the release and vote; the vote will run for the usual 5 days,
ending on 03/22/2017 PDT time.
Thanks,
Junping
---------------------------------------------------------------------
To unsubscribe, e-mail: hdfs-dev-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-dev-h...@hadoop.apache.org
