I came across this while trying to find a solution for my load-balancing clustering issue in another thread. Since I perceive this as something on its own, I decided to post a new thread.

I came across this site to setup a two-server cluster using Conga - Luci and Ricci.

From my understanding, this is what was done:-
1. Both servers have luci running using a shared storage via GFS to provide a failover
2. Both (and same) servers have ricci agent running for the clustering

I'm aware that Conga clustering aim to achieve High-Availability.
My question is:-

1. Can it also achieve Load-Balancing?
2. If yes, would the guide from the link above be sufficient or would they be additional setup/pre-req to be done? Since I'm new to this area and there was no mention of load-balancing, I'm still wondering if this is what I need to incorporate.

Thanks in advance.

My experience with GFS(2), Conga, Luci, and Ricci is fairly limited, but I don't think that load balancing in the way you mean is it's intent. Load-Balancing requires some agent to be aware of the state of some nodes, and then hand off the load evenly between the available nodes. The Cluster-Linux folks are aimed at a decentralized cluster, where there is no such thing as a "master" node. Everything is negotiated between the nodes to figure out what should happen, which is why GFS(2) works the way it does.

But to be a little more direct, what service are you intending to load balance? Are you wanting to load balance a website via apache? Or NFS traffic between the cluster nodes?

Lastly, I would suggest you read up on the official Redhat Documentation, as they are intensely interested in the Linux-Cluster software, and include versions in the most recent RHEL 5.5 release.

http://www.redhat.com/docs/en-US/Red...iew/index.html

Hi Arch

Thanks for the reply. Already resigned to the fact that this thread is a dead one.

Well, your explanations does help. As for your question, the situation is this:-

The original design utilizes a single server (lets call it A), and many other servers will routinely (e.g. 15mins interval) perform a RCP to server A to transmit data over. Since the "many other servers" can number up to 200+, and a preliminary testing I did with 250 concurrent RCPs incoming, another server B was decided to be introduced to cluster and load-balance with server A.

This is the original thread, but its not Luci-Ricci specific.

I'll be sure to check up on the site. Thanks again.

ya, the load balancing piece isn't something that the linux-cluster software is built to handle. You could think of it more as the piece that handles the load balanced traffic, but it doesn't to it itself.

You could use something like Round-Robin-DNS, which isn't a great way to load balance, since it's DNS based, and DNS is cached, and the only metric used is when DNS is requested.

So you'd setup a hostname for 'server.com' that has two A, or two CNAME records that point to server A and server B. Then when you need to add a third server into the mix, it's a matter of adding the new server to the cluster, and setting up a new DNS entry.

Hope this helps

I answered your post because LQ is doing a "Zero Reply" drive, and I thought it would be interesting to see what sort of questions don't get replies. I was surprised to find that there were a handful I had some experience with that were rather old, so I figured I'd give 'em a shot.