See https://builds.apache.org/job/Hadoop-Common-trunk/1039/changes
Changes:
[sandy] Fix YARN-1697 in CHANGES.txt
[shv] HDFS-4858. HDFS DataNode to NameNode RPC should timeout. Contributed by
Henry Wang.
[zjshen] YARN-1719. Fixed the root path related Jersey warnings produced in
+1 (non-binding)
Built source. Brought up a pseudo-cluster with manual RM HA. Ran a few MR
examples.
On Tue, Feb 11, 2014 at 10:40 PM, Roman Shaposhnik r...@apache.org wrote:
Very nice! I'll take it for the usual round of testing in Bigtop.
Should take a few days, but I'll keep you guys
+1 (non-binding)
- Upgraded HDFS including snapshots from 2.2.
- Ran a few sample jobs.
- WebUI looks good.
Thanks,
Akira
(2014/02/12 7:09), Karthik Kambatla wrote:
+1 (non-binding)
Built source. Brought up a pseudo-cluster with manual RM HA. Ran a few MR
examples.
On Tue, Feb 11, 2014
Larry McCay created HADOOP-10342:
Summary: Extend UserGroupInformation to return a UGI given a
preauthenticated kerberos Subject
Key: HADOOP-10342
URL: https://issues.apache.org/jira/browse/HADOOP-10342
Running pseudo cluster out of the box (expanding the binary tar, or
building from source) does not work, you have to go an set the MR framework
to yarn, the default FS URI to hdfs://localhost:8020, and so on.
While I don't see this as showstopper (for the knowledgeable user), it will
make may
+1 (non-binding)
I downloaded the source tarball,
built it,
started pseudo distrubuted cluster,
run wordcount example job.
The RC is available at: http://people.apache.org/~acmurthy/hadoop-2.3.0-rc0
Though the hadoop-2.3.0.tar.gz downloaded from above URL
contains site documentation in
I don't think any of what you describe below is a regression in behavior
from earlier releases. The fs.defaultFS has been set to file:/// for a long
time, and likewise you've similarly had to set up your YARN configs. Given
that, I don't think this warrants a new RC.
--
Aaron T. Myers
Software
+1 non-binding
- checked out the rc tag and built from source
- deployed a pseudo-distributed cluster with
yarn.resourcemanager.recovery.enabled=true
- ran a sleep job with multiple map waves and a long reducer
-- SIGKILL'd AM at various points and verified AM restart
-- SIGKILL'd RM at various
+1 (non-binding)
downloaded the source tar ball, built, ran a number of MR jobs on a
single-node cluster and checked the job history from job history server.
On Wed, Feb 12, 2014 at 7:53 PM, Gera Shegalov g...@shegalov.com wrote:
+1 non-binding
- checked out the rc tag and built from source