(Simple?) High availability question

[Ansar Mohammed]

There are many types of HA solutions.
1. Load balancing: This assumes that there is little or no data to
synchronize and basically the client connections are load balanced between
two or more hosts. Eg: DNSRR, OpenBSD pf, Cisco  and WLBS. Typically this is
used to load balance web servers, dns servers and firewalls.

2. Shared Nothing: This consists of two or more physical servers connected
back to shared LUN on scsi bus, FC SAN or iSCSI. In the Windows and UNIX
worlds this is what "clustering" refers to MOST of the time. You can do the
same thing on Linux/FreeBSD using scripts. This is typically how MOST
database servers are clustered.

3. Application Clusters: Application specific replication technology that
replicates data and manages client connectivity. Eg: Active Directory,
Oracle RAC, SQL Server. It can be replicated in an opensource environment
with rsync and DNS

4. Distributed systems: many systems working together distributing
processing among them eg: Beowulf, PVMs, Windows Compute Cluster.

To solve your problem for high availability you should ideally use a shared
nothing cluster for the database and DNS RR or IP Load Balancing for the web
servers.



> -----Original Message-----
> From: owner-tlug-lxSQFCZeNF4 at public.gmane.org [mailto:owner-tlug-lxSQFCZeNF4 at public.gmane.org] On Behalf Of Madison
> Kelly
> Sent: June 1, 2007 6:00 PM
> To: tlug-lxSQFCZeNF4 at public.gmane.org
> Subject: Re: [TLUG]: (Simple?) High availability question
> 
> Ian Petersen wrote:
> > On 6/1/07, Lennart Sorensen <lsorense-1wCw9BSqJbv44Nm34jS7GywD8/FfD2ys at public.gmane.org> wrote:
> >> I believe mysql has a replicating
> >> server backend, although apparently that backend is much slower and has
> >> less features than the regular one, so it is a major tradeoff there.
> >
> > I used to work for a place that relied on MySQL for the majority of
> > its database needs.  I believe the MySQL dbs were all replicating in
> > multi-master mode.  It was apparently adequate for their needs, but I
> > don't know what happens if/when one of the db machines fails.  I've
> > heard that multi-master database replication is a thorny problem from
> > a theoretical perspective (I think it was Seneca and Christopher
> > Browne I overheard at a LUG meeting).  I also remember that the tech
> > support guru occasionally had to issue strange commands to get things
> > unwedged that included a memorized magic byte offset into MySQL's
> > binary logs.  On the other hand, this company had 24x7 uptime
> > obligations to its customers and unscheduled downtime had direct
> > consequences on their bottom line, so the replication must have been
> > doing something worthwhile.
> >
> > I prefer PostgreSQL for my own database needs, but I think it's
> > because I was exposed to PostgreSQL before MySQL, and, at the time,
> > MySQL's position on ACID was reprehensible.  I hear that MySQL 5 is
> > much better ACID-wise, and it's had replication for a long time.  It
> > might be worth looking into.
> >
> > Ian
> 
> Well, I use roundcube webmail with MySQL holding it's data (couldn't get
> it working on PgSQL). Otherwise I wouldn't have to worry about MySQL at
> all. In this case though, I do, so thank you for the info. I hadn't
> asked about MySQL yet because I was trying to keep my poor noggin'
> focused on as few crazy things as possible. In our case though,
> switching to MySQL is both not appealing and not practical. It's nice to
> know it'll work for the few instances we need it though!
> 
> Madi
> 
> 
> --
> The Toronto Linux Users Group.      Meetings: http://gtalug.org/
> TLUG requests: Linux topics, No HTML, wrap text below 80 columns
> How to UNSUBSCRIBE: http://gtalug.org/wiki/Mailing_lists

--
The Toronto Linux Users Group.      Meetings: http://gtalug.org/
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://gtalug.org/wiki/Mailing_lists