|
Stack |
at Dec 22, 2010 at 7:14 pm
|
⇧ |
| |
(I moved this topic to
[email protected] from user).
On Wed, Dec 22, 2010 at 10:12 AM, Gary Helmling wrote:
What you have there currently seems accurate. So I don't think it needs to
mention HBASE-3194 directly. Maybe add a note that we try to support Hadoop
0.20.x variants incorporating security features as well (CDH3B3 and Y!
Hadoop 0.20.S)? Done.
From a user standpoint it does seem a bit complicated though and doesn't
help the criticism that we have "too many moving parts".
...
But that would be akin to naming a "preferred" Hadoop distribution for the
project (we may all have our own preferences anyway), which doesn't seem
like a place we're at yet.
Well, the 'preferred' is the Apache distribution. Thats what we
should bundle, etc.
That said, I get that you were trying to talk more to our current
predicament where our setup is more involved than it should be for new
users because the 'preferred' Hadoop is a source only distribution and
many will be running other than the Apache distribution anyways. In
general we need to do work -- documentation, introspecting shims, etc
-- to make it so HBase is more 'universal' and can more easily be
deployed atop any Hadoop whether Cloudera's CDH, newer versions of
Hadoop (hadoop 0.21, 0.22, 0.20s), or vendor X's implementation of
HDFS, Ceph, etc.
I like the idea of linking off to Cloudera doc. -- or any other
vendor's doc -- from ours for install (Todd?). Or, maybe better would
be linking to a wiki page that vendors can edit as they wish?
Let me ask Dhruba what he thinks about making a 0.20-append release
(He's the release manager). Will also sound out the hadoop pmc since
they'll have an opinion.
St.Ack