On 07/09/2013 09:47 AM, Sean Mackrory wrote:
Without wanting to detract from the spirit of focussing on system
stability, I'd like to suggest a few changes I think it's time we at least
discuss seriously:
JDKs: I've seen a lot of people ask about JDK 7. Perhaps time to add
support for Oracle JDK 7? It's working pretty well in my experience, and
although it's less tested upstream, the only JDK we officially support is
officially EOL, so we're not exactly in a good position now IMO.
Debian 7 has also been out for a while, and I think we should do at least
one release on it. It's likely very little work but I think there's value
in certifying the stack will work well there. (On the topic of OS's - are
we specifically talking SP3 of SLES 11?). I don't feel strongly on this,
but I'm just curious if there's a reason you're suggesting staying with
12.10 and not 13.04 - other than wanting less change in this release?
Again, I hardly have an opinion on that one.
Other components that have recently had releases that I don't consider to
impact the
- Hue 2.4.0
- Whirr 0.8.2
- Flume 1.4.0
There's also been a ticket to package Avro for a long time and I'd like to
get to that soon. Perhaps Parquet as well? Although like Phoenix and
DataFu, I would suggest doing just the libraries for now, not all the CLI
tools.
Again - I don't mean to take away from the focus on stability, but I also
don't think we shouldn't stretch to stay up to date either.
+1 to everything else as suggested, however.
On Tue, Jul 9, 2013 at 8:59 AM, Andrew Purtell <[email protected]> wrote:
Would you be willing to consider Phoenix, only BIGTOP-993? Installing the
package produced by 993 only drops a library for HBase into
/usr/lib/phoenix, essentially the same relationship between the DataFu
package and Pig. There is follow up work that is more ambitious, for
example BIGTOP-1007, but that is not required by any means and could come
in if/whenever you are comfortable with it.
On Mon, Jul 8, 2013 at 8:50 PM, Konstantin Boudnik <[email protected]> wrote:
Guys,
I wanna kick-off the discussion on the content of 0.7.0 BOM
Release 0.6.0 was all about stabilization of the stack and I think we
got a
great headway on that. The following components/OS were in the frame of
the
discussion:
http://is.gd/H52iVe
My personal take that we need to spend this release cycle working on the
improvements in Bigtop itself: we got enough "technical debts" in the
pipeline that have to be addressed. To name a few:
- testability/test coverage
- test framework
- package improvements
- build improvements (including performance)
In order to be able to deliver a solid stack again yet improve all things
Bigtop I'd like to focus on the latter, hence keeping the former at bay
and
limiting the component updates to the bugfix releases only (if
warranted).
E.g.
Hadoop 2.0.5 or later (stabilization branch of Hadoop 2)
HBase 0.94.9 (update from the Bigtop 0.6.0)
HCatalog 0.5.0 (same as 0.6.0, as well as following...)
Zookeeper 3.4.5
Pig 0.11.1
Hive 0.10.0
Sqoop 2
Oozie 3.3.2
Whirr 0.8.1
Mahout 0.7
Flume 1.3.1
Giraph 0.2.0
Hue 2.2.0
Datafu 0.0.6
Solr 4.2.1
Crunch 0.5.0
Tomcat 6.0.36
Spark 0.7.3 (it has been in the queue for a long time)
Also, I'd suggest to keep the same set of OSes as last time:
CentOS/RHEL5
CentOS/RHEL6
SLES11
Ubuntu 12.04 (LTS)
Fedora 18
OpenSUSE 12.3
Ubuntu 12.10
To reiterate, with a known stable version of the stack we can safely
focus
on
the improvements to the framework and the overall system usability.
Please jump on the discussion Also, I have opened up the following JIRA
to
track the BOM update https://issues.apache.org/jira/browse/BIGTOP-1023
Thanks,
Cos
--
Best regards,
- Andy
Problems worthy of attack prove their worth by hitting back. - Piet Hein
(via Tom White)
+1 to what Andrew and Sean said.
I would also like to add:
* I would also like to keep the door open for Apache Gora to come in.
Someone is already working on it
* Fedora 19 is out and maybe 20 will be out by the time we release 0.7.0
With regard to stability, if included all the changes listed so far seem
a little bit too much, what about dedicating a 0.6.1 to stability so
0.7.0 can include all the shiny things listed above? A 0.6.1 would
enable us to be more strict with the patches/fixes/updates being
included as well as signaling a more stable version to users.
Thanks,
Bruno
