we are running a web portal that serve huge number of the visitors, and it seems that we are running out of resources on the server, so I have a question about clustering.
Is it possible to run 3 or more servers at the same time (run Virtualmin on all of them) so they act as one server, share cpu, ram-mem and hdd ?? is this possible with Virtualmin ? and if Yes how do we run multiple servers behind the same real wan ip address that our domain points to … ?
In short words we need around 3-4 servers to act as one server (all of them in the same cluster where they share CPU, RAM and HDD…)
If this is possible than I assume that all of the servers must run the same hardware and same configuration as well ?
That’s a very broad question without any details, but I don’t think there is any one tool that can do what you want out of the box. It really depends on the application.
In any case this would be a “very hard work” scenario AFAIK (but I am not really an expert in the area of high availability).
But to make some steps easier in this scenario you could look into Cloudmin. It is very nice tool to manage multiple physical servers and/or (all types of) virtual machines (e.g. aws).
it`s the Apache, there is large amount of the connections… if You have any idea of tuning the Apache as the “best practic” I would like to know …
The CPU and RAM in it self are not the “core” problem, but we will add more RAM to the server.
But anyway as You already mentioned there is a lot of ways to “load balance” a server and none of them will work just out-of-the-box …
One of them is via DNS round Robin but the core problem here is if one of the servers goes down, than users who used the ip address of that server will not reach it because this server is down and the user does not know the new or another ip address for that domain to reach the another server…
Another option is via “Linux HA Heartbeat” where both servers have its own ip address and one virtual address which is used by Heartbeat, and than both servers or “multiple” servers are hidden behind one WAN ip address…
This is also good idea, but as far as I understand the Heartbeat works in the following way, it will always use the primary server and not secondary, secondary is only used if the primary server fails, or if the primary “Apache server service” crashes … as Heartbeat is the HA that is used by the Apache in the most cases…
So I am looking for a solution to run 2 or more servers behind only one WAN ip address… BUT I will load balance between them, which means, I would like to send one request to to one of the servers, another request to another server and third request to the third server etc…
I hope that someone who is running HA between the CentOS server can post some useful explanation…
