couchdb-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Filipe David Manana <fdman...@gmail.com>
Subject Re: Anyone ETA on fixing push replication?
Date Thu, 10 Jun 2010 18:22:33 GMT
By curiosity, are you using CouchDB 0.11 or trunk?
That problem should be gone with trunk.


On Thu, Jun 10, 2010 at 6:19 PM, Bharat Bharat <Bharat.Bharat@citrix.com>wrote:

> Thanks Chris.
>
> On Jun 9, 2010, at 9:33 PM, J Chris Anderson wrote:
>
> >
> > On Jun 9, 2010, at 8:41 PM, Bharat Bharat wrote:
> >
> >> I find push replication very very buggy for 2 reasons:
> >>
> >> - Very high failure rate for replication across different subnet.
> >> - Inflates target DB size to almost like 10 times.
> >>
> >> For my exact test scenarios, pull replication works flawlessly and push
> fails like almost every time. Couchdb itself recommends pull replication
> over push. This however leads to another problem that you have to log in to
> your target DB instances to trigger pull(or use curl to specify source and
> target urls). I think the problem starts when you try to specify a URL
> instead of just DB name in the target body.
> >>
> >> curl -X POST -d '{"source":"db","target":"http://10.1.1.1:5984/db2"}'
> http://localhost:5984/_replicate
> >>
> >> I was wondering if there is some sort of fix/patch out there or any plan
> to fix this in upcoming releases?
> >>
> >> Thanks very much
> >>
> >
> >
> > Yes, these problems are well known. There's been lots of work going into
> the replication APIs for the next release.
> >
> > You might try trunk and see if it's better for you.
> >
> > And Damien is working on a brand new replicator, which should be much
> simpler code, so for 1.1 or some future release we'll have licked this stuff
> for good.
> >
> > Chris
>
>


-- 
Filipe David Manana,
fdmanana@gmail.com

"Reasonable men adapt themselves to the world.
Unreasonable men adapt the world to themselves.
That's why all progress depends on unreasonable men."

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