hbase-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Stack <st...@duboce.net>
Subject Re: [UPDATE] Finishing up 0.96 --> WAS Re: 0.95 and 0.96 remaining issues
Date Thu, 15 Aug 2013 15:21:18 GMT
On Wed, Aug 14, 2013 at 11:44 PM, ramkrishna vasudevan <
ramkrishna.s.vasudevan@gmail.com> wrote:

> Hi Stack
> First I would like to thank you for the reviews and understand the
> decision.
> I would just like to highlight few facts if this does not make into 0.96
> then we may have to
> -> Handle backward compatability issues with WAL.  Currently inorder to
> retrieve tags we write a taglength thought it is 0.  This would help us to
> replay WAL with Tags.

Is there not a version in WALEdit or on the WAL file that would allow us
distinguish a WALEdit written w/ tags from one w/o?

> -> Codecs like PRefixTreeCodec needs to be working for versions without
> Tags.  Currently as PrefixTreeCodec is not out in any release these are not
> a problem for now.

Could we have two codecs, one that supports tags and another that does not?

> So to avoid these problems we can get these Tags on 0.96 release.  May be a
> few more days of time.  Is it possible?  I can try getting people for more
> reviews too.

My sense going by recent experience is that landing big patches, what w/
review cycles and then stabilization of build post commit, is that the
process always takes way more time than I expect.  You fellas are almost
there -- tags is shaping up nicely -- and you and Anoop are super receptive
and reactive to criticism so it would all likely run faster than usual but
I predict at least another week before it lands -- perhaps too weeks -- and
that is too long for a 0.96 that is already way too late.

Lets get tags into trunk. I can help after this development release and the
release candidate gets posted.  Lets then make a tags-only release on the
heels of 0.96.  As said above, tags is cause enough for a major hbase

Good on you Ram,

  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message