I am trying to find a solution that will allow me to start with two basic Linux servers that would be running Linux, Apache, mySQL, PHP, Samba and NTP. Then as need be more servers could be added in a hot-plug fashion to give more power and speed as the users increase.

Ideally the system should function as long as only one server works and the more servers that are connected the higher the availability/processing power will be. The other big item is that disk space needs to be configured so additional storage can be added. Either separately with something like NAS or hot-plug server RAID solutions. Linux though needs to be able to “grow” the file system to use additional storage as it is added.

I have been looking around for awhile and can not find in-depth information about this. I’ve been reading web sites about Linux clustering, Beowulf clusters, heartbeat, OpenMosix, and the Linux Virtual Server Project. After considering buying the Linux Clustering book it seems like it might do more harm then good with all of the bad reviews.

Is what I’m talking about possible? (I’m sure it must be) Then my main question is where can I find some type of documentation that can walk me through how to set something like this up? At this point crumbs would be appreciated, but a walkthrough/howto would be like gold.

Also please note that I’m looking for something to run on basic hardware, not old 486 stuff, but basic new systems. Also as a ballpark figure there will be about 150 web users and 10 file transfers (Samba) going on at once. So I believe I’m looking at basic interconnects such as Ethernet and not specialized equipment. Also in regards to storage we are looking for cheap but effective. So SANs would probably be out.

--
Thanks in advance
Blaine

I don't know how well the support is for the controllers are in Linux, but Dell's perc cards in combination with a PowerVault may be a good option for you. You can get a basic PowerVault for a little over $2,000 (plus the perc cards and cables which will be another few hundred each - plus drives). Just one thing though: throw their instruction books away because to cluster using a PowerVault their instruction manuals, at least of September 2004, are COMPLETELY wrong.

Sounds like you'd want to setup some type of load-balancing + failover cluster rather than a true parallel processing cluster like a beowolf. Beowolf and other HPC clusters are designed more for performing intensive parallel computations and don't give you much of a advantage when dealing with tasks like serving a web page or accessing a SMB share (you can't really divide up either of those tasks across multiple nodes/processors very efficiently).

OpenMOSIX has some load-balancing features, but I'm not sure how it stacks up to other true load-balancing designs like LVS. If you do setup a load-balancing or round-robin cluster, make sure to connect all the front ends and backend DB stuff together using high bandwidth connections to reduce communication bottlenecks. 100Mb should be alright, don't need anything ludicrous like Infiniband.

I don't know how basic your LAMP server setup is going to be but for in depth information on the Server apps that you mention I would suggest that you read "Managing Linux Systems with Webmin" by Jamie Cameron.

I am reading this book now and there are about three chapters on clustering servers, users and services.

Good luck with your project

Thorn

Try to read info http://www.redhat.com/docs/manuals/csgfs/ . IT will help you.

I think this should explain everything you need to know:
http://www.linuxvirtualserver.org/