hadoop-mapreduce-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Konstantin Shvachko <shv.had...@gmail.com>
Subject Re: VOTE: Committing HADOOP-6949 to 0.22 branch
Date Wed, 30 Mar 2011 01:08:15 GMT
Nigel,
The nature of incompatibility is that the RPC version is changing, which
means
VersionedProtocol-s become incompatible all at once. As opposed to say
only DatanodeProtocol or mr.ClientProtocol.

Doug is right because of our strict requirements for protocol compatibility
this
will not affect users upgrading to 0.22.

Matt, thanks for checking and testing.

--Konstantin

On Tue, Mar 29, 2011 at 11:04 AM, Jakob Homan <jghoman@gmail.com> wrote:

> +1
> n.b. that the vote lost hdfs and common dev at some point.  I've added
> them back.
>
> On Tue, Mar 29, 2011 at 9:18 AM, Amit Sangroya <sangroyaamit@gmail.com>
> wrote:
> > +1
> >
> > On Tue, Mar 29, 2011 at 6:04 PM, Stephen Boesch <javadba@gmail.com>
> wrote:
> >> +1
> >>
> >> 2011/3/29 Doug Cutting <cutting@apache.org>
> >>
> >>> +1
> >>>
> >>> I don't think this creates an incompatibility.  It changes the RPC wire
> >>> format, but we already require that clients and servers run identical
> >>> builds.  No application that ran with a prior version of Hadoop would
> be
> >>> broken by this change when it upgrades to this version of Hadoop.
> >>>
> >>> Doug
> >>>
> >>> On 03/28/2011 09:39 PM, Konstantin Shvachko wrote:
> >>> > HADOOP-6949 introduced a very important optimization to the RPC
> layer.
> >>> Based
> >>> > on the benchmarks presented in HDFS-1583 this provides an order of
> >>> magnitude
> >>> > improvement of current RPC implementation.
> >>> > RPC is a common component of Hadoop projects. Many of them should
> benefit
> >>> > from this change. But since this is an incompatible change it
> requires a
> >>> > vote to be included into a previous branch.
> >>> > Please vote for inclusion of this change into branch 0.22.
> >>> >
> >>> > +1 from me.
> >>> >
> >>> > Thanks,
> >>> > --Konstantin
> >>> >
> >>>
> >>
> >
>

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