Hi, I'm setting up an ISPconfig multiserver setup on AWS and I'm currently running into troubles. So I wonder if the concept can be realized the way I want it. My plan is: run ISPconfig on EC2 instances within an AutoScaling group using a load balancer have the DB on RDS (separate database server) have the userdata (var/www) on EFS (network share) What does that mean: Within the AutoScaling group, new instances are started whenever an instance dies or the total load is above a limit. Of course a new instance gets a new IP and hostname. From the outside world only the load balancer is reached via a fixed IP. It routes the traffic to a specific instance. The database and the network share are used by all instances, so data is consistent all the time. The instances have an internal IP/hostname for communication within the VPC and an external IP/hostname to be reached from the internet. The new instances are not dedicated to a specific task, so they should do the same things as the first instance. The database on RDS has some limitations. Especially it doesn't allow to grant ALL PRIVILEGES. That makes the installation difficult, since manual steps are necessary. Questions: Can ISPConfig deal with instances coming up with new IP/hostname? Which IP/hostname (internal/external) to I need to use in the specific installation steps/config files? Can ISPconfig cope with the load balancer holding the only fixed IP, or do I need a fixed IP for every instance? Is it possible to have multiple servers with equal functionality, or does ISPconfig need dedication? Is it possible to share userdata on a network share, or does it confuse ISPconfig? Is there a way to do script based installation without the script trying to grant ALL PRIVILEGES to a new user? These are only a few of the questions, that arouse during installation. But I think these will help me decide whether to continue on ths approach or change the concept and start from scratch. Thanks, Stefan
1: It should be possible, but I think you will always need some manual configuration. 2: Both can be used. For my multiserver setup I use the local IP's so local traffic is not calculated in my traffic limits. 3: I think this should work, but it's best practice to have a (eventually local) fixed IP for communication between slaves and master and eventually between slaves 4: You can set servers to mirror one of your other servers, e.g. you have web1, web2, and web3, web2 and web3 could be set as mirror of web1. 5: This is possible, use fstab to mount it (no symlinks) 6: I don't think so. Maybe someone else has something like this working and can share their experiences, but I think it will be quite complicated.
Thanks for your fast reply. Yes, it's quite complicated, especially when you are quite new to both aws and ISPconfig, like me. But I took a few great learnings out of that so far, and I'm not finished yet... Your answer tells me, thats it is not completely hopeless, but still you leave a lot of questions open. I would be grateful, if someone could give me some hints, especially about the database setup, which obviously has to be done manually. thx, Stefan
