I not understadn 100% how clusterd servers work.
My basic questions are:
Whats up if i have clustered boxes, 2 boxes, and one die? all continue working?
Whats up if some site its hacked, that hack, propagate to all clusters?
Whats up if some files its demaged due hard disk failiure in one of theclusters, that demaged file propagate to the rest?
One have cluster, imagine have 2 boxes, one is the master box amange load balancing and if that box down, all sites and services go down?
I apreciate if somebody can epcxlain that, i understand the major goal in cluster balanced servers its disribute the load around multiople boxes…but i wnat knwo more about how much reliable its a clustered system, related to hard disk carshes, or DoS attacks, or hardware failures on obe of that boxes.
One acts as a Master server, to load balance requests. Another acts as a second mirrored Slave. The third acts as file storage.
If they’re set up as Primary and Secondary DNS, if the first one dies, the second one will take up the DNS requests and work properly. If the second one dies, the first one will take up the slack. However, AFAIK, MySQL runs on the Primary server — MySQL sites on the second server will die.
Since all files are stored on an NFS mount on the third server (which doesn’t have to be directly on the internet) then if one machine gets hacked by, say, a SQL injection bug or a malicious script that rewrites the homepage, then of course it’s going to affect both servers.
At least, that’s the way I understand it … if I’m wrong, please correct me.
There is currently only one CM, and all the other boxes in the cluster are nodes. The CM acts as the file server, the load balancer, and the MySQL server. All the nodes mount, via NFS, /chroot/home on the CM.
All traffic to sites using virtual IPs (IPs available to the cluster) always hits the CM first, at which point the CM will route the traffic to the correct node. The node will then handle the request and send the data back to the visitor/client. The next subsequent request from the same visitor/client still goes to the CM first. It is the load balancing policy on the CM that determines if the same node gets that request.
The current clustering setup in InterWorx is not high availablity. Say you 4 boxes in your cluster and you have a simple round-robin load balancing policy (evenly distribute traffic among all nodes, regardless of how many connections each node is currently handling). If one of the nodes go down, 1 out of 4 times, the site will appear offline. The load balancer currently can not tell if a node is down and does not adjust accordingly. Additionally, if the CM goes down, the entire cluster goes down since the CM is the load balancer.
If the hack affects files in the /home directory, then yes, all boxes in the cluster will be affected because they all share the /home directory. However, each box in the cluster has it’s own /var/tmp and /tmp directory, so there is a little localization there. In other words, the malicious script may have targeted files shared by all the boxes, but the script could be installed only on one of the nodes.
If the CM’s hard drive fails, all boxes are affected. If a node’s hard drive fails, that won’t cause the drives on other boxes to fail, but the node will be down causing disruption of service as I mentioned above.
Exelent repply : ) now understand… work exactly like i think,… clustering act like one single machine… the goal of use cluster is distribute the traffic, example, monster FORUM site with very high traffic, where one single box cant deal with all that traffic and apache overload the sever, clustering its great solution for situations like that.
For virtualhosting. (my situation), i think its not really usefull, or can be if have to deal hosting sites with very high traffic and lot of request to apache (not my situation)
For standar hosting i think just add complications.
