LVS
lvs-users
Google
 
Web LinuxVirtualServer.org

RE: LVS-DR Filesystem

To: "'LinuxVirtualServer.org users mailing list.'" <lvs-users@xxxxxxxxxxxxxxxxxxxxxx>
Subject: RE: LVS-DR Filesystem
From: <Webmaster@xxxxxxxxxxxxxxxxxxx>
Date: Thu, 16 Feb 2006 14:13:55 -0500
Sorry, I should have gone into more detail about the set-up.  

I want to set up an LVS-DR cluster using purely Red Hat Enterprise Linux 4
for use in web hosting.  

The services I would need to run would be Apache, MYSQL, BIND, PHP, etc.
Anything you could think of for use in serving web sites.  

What do large web server farms run?  What do big companies use for this?
I'm very familiar with the book: "Linux Enterprise Cluster: Build a Highly
Available Cluster with Commodity Hardware and Free Software" by Karl Kopper.

In the book, he uses a NAS box to handle the MYSQL and FS.  I just don't
like the idea of using a single box.  Isn't there anyway you can utilize
each hard drive on each real server node, kind of like a virtual raid array?



-----Original Message-----
From: lvs-users-bounces@xxxxxxxxxxxxxxxxxxxxxx
[mailto:lvs-users-bounces@xxxxxxxxxxxxxxxxxxxxxx] On Behalf Of Graeme Fowler
Sent: Thursday, February 16, 2006 3:20 AM
To: LinuxVirtualServer.org users mailing list.
Subject: Re: LVS-DR Filesystem

Hi

On Thu, 2006-02-16 at 02:57 -0500, Webmaster@xxxxxxxxxxxxxxxxxxx wrote:
> What is the best way to handle a filesystem using an LVS-DR cluster?  

Do you really mean "what's the best way to have a shared filesystem used
by my realservers"?

> I have heard of using a NAS box, but from what I understand it is only
> recommended to use a max of 8 nodes with that.

Well, that depends on what service you're providing and what sort of
budget you have. I'm using a NetApp FAS270c in one production platform
to provide both NFS and CIFS (SMB) exports/shares to <counts...> about
26 mixed Windows and Linux realservers (web and mail), and it isn't
breaking a sweat. It will scale much further than it already has.

> I have also heard of GFS, DFS, AFS, etc.  I want to make the cluster
highly
> scalable and would like to not have a ceiling on the max number of nodes.
> If I use something like say GFS, will I have any data corruption problems?


I haven't much experience of these sorts of filesystems, but IIRC the
last time I read RedHat's whitepapers about GFS it required some sort of
consolidated backend like a SAN which you then export LUNs from towards
your GFS cluster. It wasn't a road I wanted to travel down, so I went
with trusty NFS - its' foibles are well documented.

I'd say you should get yourself a powerful, resilient, multi-cpu and
multi-psu server with a good RAID array using fast SCSI disks (15k RPM)
and then use it as an NFS server. Of course, you haven't explained
entirely what it is you're doing, so this could be way off the mark :)

Graeme



<Prev in Thread] Current Thread [Next in Thread>