On Saturday 12 November 2005 06:50 pm, Dormando wrote:
[snip]
>
> As far as seamless redundancy goes; http://www.drbd.org/ is probably
> your tool, along with MySQL replication if desired. Set up any mounts
> you require exist on both systems onto one or more drbd partitions,
> create your filesystem on top of that, then set up a keepalived script
> (along with ucarp?) to bring up the backup volume when the primary is down.

I've used DRBD a lot and it works well, but you can only have one system at a 
time mount the drbd devices. The original poster was looking to have a 
active/active cluster, which drbd does not support.

Active/active clusters are hard. I'd recommend starting with an active/passive 
cluster first.

You will need at least 3 ip addresses. One ip for each box and one 'floating' 
ip for the services (like http, smtp, pop, etc). Having the boxes at two 
separate ISPs is impossible since you won't be able to get them to fail over 
the 'floating' ip address, unless they really know what they are doing (both 
of them).

P.S. I've had problems in the past with the -gentoo kernels and drbd. Oh, also 
4k stacks seems to give drbd problems. But it's been a while since I've 
upgraded the drbd software on the boxes running it, so that may be fixed now.

>
> It's not free, it's not going to set itself up for you, but it's the
> most relatively uncomplicated setup I can mention at this point.
>
> -Dormando
>
> Eric Brown wrote:
> > Some food for thought:
> >
> > net-misc/ucarp for redundant IP configurations.. maybe you could have
> > one box use the original IP if the first one fails, while also having
> > its own primary IP for concurrent action...
> >
> > For data redundancy, I've only really heard of stuff like AFS where
> > you could have lots of replication.  But if you don't need immediate
> > replication, you might be able to get away with hourly rsyncs or
> > something..
> >
> > What I don't like about the Linux Virtual Server model (having one
> > machine distribute the load to real-servers) is that it's just another
> > form of weakness.  You could use ucarp to provide redundancy on the
> > distributing machine, but it all seems too complicated to work well.
> > Anyway, if you're considering going that far, you might want to check
> > out openMosix for the apache process distribution.
> >
> >
> > Personally, I would want to try having multiple MX records, and
> > perhaps some simple http request distribution too (maybe DNS can
> > manage something like that).. All of the data could be rsync'd and
> > mail could possibly be forwarded between the two machines to keep
> > stuff up to date (that, or use some slow AFS/NFS implementation..)..
> > Or consider some kind of database for storage...
> >
> > Eric
> >
> > On 11/12/05, xyon <[EMAIL PROTECTED]> wrote:
> >>But what kind of solution for config files and mail? I'm running into
> >>the same dilemma with my setup. In my previous setup I had one server
> >>NFS-mount the other server for mail and things, but that just does not
> >>accomplish the data redundancy that I need.
> >>
> >>Mal, as far as database setup, I would check out setting up MySQL
> >>Master/Slave replication (or if you run MySQL 5.0 you can have dual
> >>Masters replicating to eachother). Here's a link for the how-to:
> >>http://forums.gentoo.org/viewtopic.php?t=241123
> >>
> >>You may want to look at doing MySQL clustering if you have access to a
> >>3rd server:
> >>http://dev.mysql.com/doc/refman/5.0/en/mysql-cluster-overview.html
> >>
> >>HTH! :)
> >>
> >>On Sat, 2005-11-12 at 20:46 +0100, Jeroen Geilman wrote:
> >>>Mal Herring wrote:
> >>>>Hi Gentoo-Server....
> >>>>
> >>>>I run a Apache2 webserver and Postfix as per the virtual mailhost
> >>>>guide...
> >>>>
> >>>>For resilience I would like to have two gentoo boxes doing these tasks
> >>>>for load balancing / redundancy - how can this be achieved?  Is there a
> >>>>tool that will allow replication of the config files / mysql stuff to
> >>>>allow the boxes to basically be a cluster ?
> >>>>
> >>>>Oh - one more thing, the boxes will sit in different data centres !
> >>>>
> >>>>Thanks in advance...
> >>>
> >>>That won't actually be all that easy to accomplish...
> >>>
> >>>The most robust solution will always be to set up a THIRD box in front
> >>>of those identical servers, which can load balance the traffic and/or
> >>>connections to both.
> >>>
> >>>Obviously, this is going to be more difficult to implement when the
> >>>servers are physically far apart.
> >>>
> >>>A very good *mail* solution would be to set up Postfix on both boxes for
> >>>the same domain, but with different MX proirities - this is a relatively
> >>>easy setup.
> >>>
> >>>For apache, you can do part of the loadbalancing yourself, but that will
> >>>require both boxes to be available at all times - meaning you don't have
> >>>a redundancy in place.
> >>>
> >>>You can implement this via DNS, but you will need flexible access to
> >>>your DNS records; with DNS round-robin you would have a certain amount
> >>>of redundancy.
> >>>
> >>>Google around, there are plenty of documents available.
> >>
> >>--
> >>[email protected] mailing list

-- 
Edward Muller - Interlix
[EMAIL PROTECTED]
417-862-0573
PGP Key: http://interlix.com/Members/edwardam/pgpkeys

Attachment: pgpahV4Yw9Xk9.pgp
Description: PGP signature

Reply via email to