-1

Sorry for a lack of jargon knowledge and experience. I've recently become a very new sys admin for my research group's cluster in my graduate program. We have a server name

bar.foo.edu

with about 10 nodes named like

bar01.foo.edu
bar02.foo.edu
...

What is the process for creating access to something like

http://bar.foo.edu/
http://bar.foo.edu/~user/index.html

On another cluster it was as simple as creating public_html/index.html with permissions that allow access, but how do you execute this setup?

physicsmichael
  • 131
  • 5

3 Answers3

1

A simple installation of Apache on the machine that bar.foo.edu is pointing to will give you the latter, but I'm not sure what bar01.foo.edu and bar02.foo.edu have to do with that... I can't imagine that you really need a 10-node cluster to do some web serving, but if you do you've got a beyond-vertical learning curve ahead of you.

womble
  • 96,255
  • 29
  • 175
  • 230
1

Well, maybe you could clarify your setup. What kind of clustering do you do? What clustering software in particular?

  • Is the whole cluster for web serving (using load balancing, for example)?
  • Do you use the cluster for scientific computations?

If it's the latter, then running a webserver on the cluster may be quite unrelated to the cluster setup. You could just run a regular webserver on one machine, then the clustering should be irrelevant. In that case womble's answer would be correct: Just configure Apache or some other webserver.

Just how to do this exactly depends on what you want to serve.

A final advice:

Why do you want to use a cluster as a web server? For various technical & administrative reasons it is usually considered preferrable to separate functions by machine. So if you already have a department webserver, student systems or similar, consider using these as the webserver; they might already be configured.

sleske
  • 10,009
  • 4
  • 34
  • 44
  • Yes, the cluster is supposed to be used for scientific computation, but no one besides me uses multiple machines at one time. I'm wondering this question because our research group lacks a representative webpage for ourselves and I hoped to hosted one at http://bar.foo.edu/ I was just wondering how to get started. Sounds like a separate machine to host is the way to go? Very low traffic, btw. – physicsmichael Sep 06 '09 at 03:27
  • In principle a separate machine is best, but for a simple low-traffic website it's ok to host it on another system. Just follow womble's answer, and post another question if there is a more specific problem. – sleske Sep 06 '09 at 16:31
0

There's probably not enough space to lay out all the possible options, but here's a couple:

  1. simple round robin'ing with bind & A records
    • bar.foo.edu has one page that 301 redirects the user to slb.foo.edu
    • slb.foo.edu is listed in DNS with 10 different A records that each point to a different server (bar01 - bar10)
    • Pros: very simple to setup; Cons: hard to manage the actual request path
  2. linux-HA
    • http://www.linux-ha.org/
    • I've used it in the past for DNS and SMTP load balancing, worked pretty well
    • Pros: more configurable than #1, increased fault tolerance; Cons: more complicated to configure
Greeblesnort
  • 1,759
  • 8
  • 10