Re: Cluster/Replication

From: "Scott Marlowe" <scott(dot)marlowe(at)gmail(dot)com>
To: "Sean Brown" <sbrown(at)eaglepress(dot)com>
Cc: pgsql-admin(at)postgresql(dot)org
Subject: Re: Cluster/Replication
Date: 2007-10-19 19:05:12
Message-ID: dcc563d10710191205y7b979cdx97a5ddf16d14dd79@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-admin

Please cc the list, as someone with a lot more experience with
pgcluster than I might be reading it. :)

On 10/19/07, Sean Brown <sbrown(at)eaglepress(dot)com> wrote:
>
> On Oct 19, 2007, at 2:30 PM, Scott Marlowe wrote:
>
> > On 10/19/07, Joshua D. Drake <jd(at)commandprompt(dot)com> wrote:
> >> On Fri, 19 Oct 2007 12:43:40 -0400
> >> Sean Brown <sbrown(at)eaglepress(dot)com> wrote:
> >>
> >>> I believe this has come up before, and I am still researching how to
> >>> do this and figured asking was probably a good idea as hopefully I
> >>> can either get some direction or someone can point me at something I
> >>> haven't seen yet.
> >>
> >> Why do you have a synchronous requirement?
> >
> > He said something about losing any data due to the loss of the master
> > being unacceptable, so synchronous was the only way to go.
> >
> > But if the machines are separated by any real distance, the speed /
> > latency of the link will be the deciding factor in the write
> > performance of the whole system.
> >
> > I think they might be better off having a local synchronous clustering
> > solution (i.e. two machines running in failover on shared storage or
> > something) and then async rep cross country if there's any distance to
> > the other server. What to do about slony not handling LOBs I don't
> > know.
>
> That can be done, the two nodes that we are first implementing will
> be in the same datacenter with a 1Gb link between them. The third
> possible node would be accessible over a 768Kb VPN but isn't being
> installed in the near future.

Yeah, that sounds like a good setup for local synch, remote asynch

> Slony is probably what we would have done if it wasn't for the BLOB
> requirement.

Well, if you're married to the BLOB, you're gonna have to find another
way to replicating. Are your blobs mostly static data? If so, you
might be able to back them up every so often and ship them across the
link to the remote machine.

> Is pgCluster as irritating to set up as it seems to be? Does the
> Replication server do a lot, does it need a lot in the way of resources?

Don't know, I haven't set it up before, only read about it.

In response to

Browse pgsql-admin by date

  From Date Subject
Next Message yogesh 2007-10-20 06:00:50 Re: Trigger AFTER UPDATE OR INSERT
Previous Message Scott Marlowe 2007-10-19 18:30:10 Re: Cluster/Replication