incubator-general mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Bruno Mahé <>
Subject Re: [VOTE] Release for Bigtop version 0.3.0-incubating
Date Mon, 12 Mar 2012 21:11:25 GMT
On 03/10/2012 07:10 PM, Roman Shaposhnik wrote:
> This is the third incubator release for Apache Bigtop, version
> 0.3.0-incubating.
> *** Please download, test, and vote by Friday, March 16
> Note that we are voting on the source (tag): release-0.3.0-incubating-RC0
> Source tarball, checksums, signature:
> The tag to be voted on:
> Please also make sure to try installing our convenience binary distribution
> artifacts. We are publishing the Bigtop 0.3.0 incubating distributions for the
> following Linux platforms: Ubuntu 10.04, CentOS 5, CentOS 6, Fedora 15,
> Fedora 16, SLES 11. The easies way to install Bigtop distribution on
> your favorite Linux OS is to pick one of the attached files, and place
> it (as root)
> in the following folder:
>    * Ubuntu -- /etc/apt/sources.list.d/
>    * CentOS5, CentOS6, Fedora 15, Fedora 16 -- /etc/yum.repos.d/
>    * SLES 11 -- /etc/zypp/repos.d/
> After that you can follow the installation instructions from over here (DO NOT
> FORGET TO SKIP STEPS #1 and #2):
> Bigtop's KEYS file, containing the PGP keys used to sign the release:
> Thanks!
> Bigtop 0.3.0 release manager,
> Roman Shaposhnik

I have started testing this release yesterday by doing a full deployment
/ tests. But I found at least an issue I would consider a blocker for
this RC: see BIGTOP-446.

I provisioned 20 CentOS 6 machines (8 core, 23GB RAM each), used our
puppet recipes to deploy this RC and tried to run some of my code on it
(importing a dump of freebase in my triplestore and query it)

Here are some notes:
1/ typo in puppet recipes. BIGTOP-446
2/ zookeeper wasn't started. Once I manually started the service
(without any issue), everything got unblocked. Probably worth opening a
3/ hbase-master couldn't start because it could not connect to zookeeper
because of 2/
4/ Somehow, the jobtracker and hdfs directories got set up only after
the zookeeper issue got resolved (because of 2/)
5/ I noticed quite a few crashes on the tasktracker/regionservers. Their
stack traces are from being helpful. But this is an upstream issue (or
me/puppet doing something stupid).
6/ Because of 5/, it would be nice to have some supervisor processes
7/ It would also be very helpful to have some
munin/collectd/ganglia/nagios integration
8/ It would be nice to be able to configure HBASE heap
9/ I would rather configure hadoop heap as a percentage of available
ram, instead of an absolute value

Overall I really enjoyed it!
The combination of puppet recipes and packages made it really easy to go
from nothing to a 20 machine hadoop cluster.


To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message