hbase-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Misty Stanley-Jones <mstanleyjo...@cloudera.com>
Subject Re: Replication terminology "master-master"
Date Fri, 12 Dec 2014 05:51:57 GMT
There is a patch and a RB request here:
https://issues.apache.org/jira/browse/HBASE-12677

On Wed, Dec 10, 2014 at 8:02 PM, Lars George <lars.george@gmail.com> wrote:

> +1 :)
>
> On Wed, Dec 10, 2014 at 5:36 AM, Nick Dimiduk <ndimiduk@gmail.com> wrote:
>
> > I suggest we drop the lingo and just say what we mean explicitly: "A
> > cluster may act as both a replication source and destination
> > simultaneously."
> >
> > On Tuesday, December 9, 2014, Misty Stanley-Jones <
> > mstanleyjones@cloudera.com> wrote:
> >
> > > Can I get some consensus on what we should be saying about different
> > > replication types? It's a bit over my head and I don't feel comfortable
> > > making a unilateral decision.
> > >
> > > On Sun, Nov 30, 2014 at 8:40 AM, Michael Segel <
> > michael_segel@hotmail.com
> > > <javascript:;>>
> > > wrote:
> > >
> > > > Both clusters are alive, but the terms active/passive means who’s
> > serving
> > > > the data.
> > > >
> > > > Your ‘master’ is the HBase cluster which is r/w and serving data.
> > > >
> > > > Your ‘slave’ is the cluster that is passive or only serving read only
> > > data.
> > > >
> > > > To your point that active/passive may be incorrect because you can
> read
> > > > from the passive, too.
> > > >
> > > > Peer to Peer may be incorrect in that while both are peers, you’re
> > > > actively writing to one, while the other gets a copy of the data and
> is
> > > > passive.
> > > > Understanding that in context a HBase cluster can be active for one
> > data
> > > > set and passive for another.
> > > >
> > > > Active/Active would be that I could write to either and there would
> be
> > > > some form of eventual consistency.
> > > >
> > > > On Nov 28, 2014, at 1:15 AM, Misty Stanley-Jones <
> > > > mstanleyjones@cloudera.com <javascript:;>> wrote:
> > > >
> > > > > I don't think active/passive is what we mean here, since both
> > clusters
> > > > are
> > > > > fully "active" from the perspective of clients, right? The slave
> > > cluster
> > > > is
> > > > > not actually read-only, is it?
> > > > >
> > > > > On Thu, Nov 27, 2014 at 10:50 PM, Michael Segel <
> > > > michael_segel@hotmail.com <javascript:;>>
> > > > > wrote:
> > > > >
> > > > >> master - master?
> > > > >>
> > > > >> Do you mean active - active ?
> > > > >>
> > > > >> replace master and slave with active and passive.
> > > > >>
> > > > >>
> > > > >> On Nov 25, 2014, at 3:26 AM, Misty Stanley-Jones <
> > > > >> mstanleyjones@cloudera.com <javascript:;>> wrote:
> > > > >>
> > > > >>> Hi all,
> > > > >>>
> > > > >>> I think "master-master" is a term that should be re-thought.
It
> is
> > > not
> > > > >>> really a "type" of replication, but refers to a characteristic
of
> > a
> > > > >>> cluster, specifically a cluster which participates in multiple
> > > clusters
> > > > >>> with different roles -- it is a slave in one cluster and
a master
> > in
> > > > >>> another cluster. I think with the current terminology, people
> > confuse
> > > > it
> > > > >>> with "cyclical" replication, in which two clusters replicate
to
> > each
> > > > >> other,
> > > > >>> and eventually each has all the data from both.
> > > > >>>
> > > > >>> Since master-master in this sense is really not a type of
> > > replication,
> > > > I
> > > > >>> think we should just scrap it. You can have master-slave
> > replication
> > > or
> > > > >>> cyclical replication, or a combination. With master-slave
> > > replication,
> > > > a
> > > > >>> cluster can fulfill both roles at the same time, as long
as it is
> > in
> > > > >>> different clusters. This is easy to understand as a sort
of
> > recursive
> > > > >>> cascade.
> > > > >>>
> > > > >>> Am I explaining it right, and what do you guys think about
> changing
> > > our
> > > > >>> terminology?
> > > > >>>
> > > > >>> Thanks,
> > > > >>> Misty
> > > > >>
> > > > >>
> > > >
> > > >
> > >
> >
>

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