On 4/1/14, 9:45 PM, Mike Drob wrote:
I did not run any tests relating to YARN. The exception sounds like
something might be wrong with your cluster set up or possibly class-path?
This shouldn't be anything required on my end as my configuration works
just fine. Assuming the same test exists in 1.5, it required no special
modification from me when I did 1.5.1.
As for the other tests, what kind of hardware are you running on? I have
pretty beefy machines that I'm using for this, so maybe things are related
to that.
8core, 32G ram. It shouldn't have any problem finishing, but this test
isn't a huge deal to me. The only concern is meeting what's stated by
the release guidelines which I believe is already satisfied (because we
don't explicitly require Apache artifacts to be used for testing).
I have no idea what the guidance on the NOTICE file is. If somebody else
has insight, that would be appreciated.
Mike
On Tue, Apr 1, 2014 at 6:11 PM, Josh Elser <[email protected]> wrote:
+0 Mostly ok, but with reservations. My votes might change to -1 depending
on whether or not the incorrect entries in NOTICE (outlined below) violate
to ASF policy (IANAL nor have I read the guidelines recently).
* Checksums/Signatures match
* Ingested and verified ~200M keys using the dist tarball on Apache Hadoop
2.2.0
* Built the source (using default profile), unit tests pass, started it
over that instance created by the dist on Apache Hadoop 2.2.0
* Ran functional tests without complete success on Apache Hadoop 2.2.0.
I'm worried that you said they passed for you -- they must pass on CDH
somehow?
Issue details
- stress.restart.RestartLoggerLate fails -- never finishes even waiting 8
times the original (fix later/never)
- stress.msplit.MetaSplitTest fails "randomly" (fix later/never)
- simple.mapreduce.MapReduceTest fails on a ClassNotFoundException on
org.apache.hadoop.yarn.exceptions.YarnException (fix later?)
- Bounced a tserver while ingesting data. After, I noticed that a log
recovery entry on the monitor stuck around for ~30 minutes until the Master
eventually logged something about removing it forcibly. Killed the tserver
again, and another log recovery entry (a second one) showed up. Not a big
deal -- I don't want more time to be wasted on the old loggers.
- NOTICE still contains references to artifacts only present in wikisearch
not in the 1.4.5 release (at least Kryo, Protobug, Minlog, Google
Collections and the enwiki file).
Mike - did you do any additional testing to verify that all of the
Hadoop1/Hadoop2 backports were caught?
On 3/31/14, 5:21 PM, Mike Drob wrote:
Devs,
Please consider the following candidate for Accumulo 1.4.5, based on git
commit f7d87b6.
Source: http://people.apache.org/~mdrob/accumulo-1.4.5-rc2/accumulo-1.4.5
-src.tar.gz
Binary:
http://people.apache.org/~mdrob/accumulo-1.4.5-rc2/
accumulo-1.4.5-dist.tar.gz
Checksums available at http://people.apache.org/~
mdrob/accumulo-1.4.5-rc2/
Staging repository:
https://repository.apache.org/content/repositories/orgapacheaccumulo-1005
All staged artifacts were build using the default profile.
Signing keys available at: http://www.apache.org/dist/accumulo/KEYS
The source tar was generated using:
VERSION=accumulo-1.4.5
COMMIT=f7d87b6
git archive --prefix=${VERSION}/ -o ../${VERSION}-src.tar.gz $COMMIT
All other artifacts were generated using src/assemble/build.sh
Testing performed:
All unit and functional tests passed using the hadoop 1 and 2 profiles.
All functional tests passed using hadoop cdh3u6 (hadoop 1 branch)
All functional tests passed using hadoop cdh4.5.0 (hadoop 2 branch)
Functional tests passed using the default profile and hadoop cdh 4.5.0
2x 24-hr CI test + verification passed (hadoop 1 and 2)
1x 36-hr CI test + verification with agitation passed (hadoop 2)
1x 72-hr CI test with agitation completed (hadoop 2)
2x 24-hr RW tests were run without agitation.
1x 24-hr RW test was run with agitation.
Changes since the previous release candidate:
Added missing license headers.
Improved the rat check used in build.sh to decrease liability of missing
licenses in the future.
Backported changes to create binary compatibility for hadoop 1/2.
Removed wikisearch issues from CHANGES.
ProblemReport needs to check for nulls on construction
bin/accumulo help should print usage
This vote will remain open for 72 hours (Until 2130 UTC, 3 Apr)
[ ] +1 - I have verified and accept these artifacts as the 1.4.5 release
of
Apache Accumulo.
[ ] +0
[ ] -1 - I do not accept these artifacts as the 1.4.5 release of Apache
Accumulo because...
Thanks,
Mike
