This is brilliant. I'm flabbergasted why every single virtual server I've ever tried has had some gotcha that made it agonizing. Why is it so hard to have one-click templates that give you a server that can run 90% of what a user needs, say apache/php/ruby/node/mysql/postgresql/ssh with cpanel (or equivalent), already installed and set up "the right way" so you can easily add domains? I know the reasons why this hasn't happened, but I don't care. I want to know the real reason it hasn't happened, and I think it has a lot to do with every problem looking like a nail. I laugh at hosting providers, for failing so spectacularly at giving customers what they want. Good on this site for finally pointing out the obvious.
>I know the reasons why this hasn't happened, but I don't care. I want to know the real reason it hasn't happened, and I think it has a lot to do with every problem looking like a nail.
That doesn't even make any sense.
Hosting services are generally not for mom and pop to start a vps on demand. For that kind of cookie cutter service, they have something called shared hosting, and its already plenty easy and has more or less what you are asking for.
If a developer can't spend couple of hours (even less unless you are a beginner) learning their way around setting up a server, it just a disaster waiting to happen.
There are time and place for cookie-cutter services, vps is not it.
Probably because SysAdmins are fervent about customization. If a simple tool pops up like this, the demands for complex features overwhelms the product and brings it all the way back to the complex nightmare it's always been.
Developers are more accepting of "Sane Defaults" as Rails has proven, but not enough hosting providers are interested in selling to developers. I have a feeling this will change sooner rather than later as the fields continue to merge.
Probably because SysAdmins are fervent about security/maintainability/reliability.
Fixed that for you. Too often all developers care about is "Ship. Ship. Ship". Get a pager and come join our world for a bit. You'll understand what its like to run a real-world, mission critical app.
use provisioning software like chef or puppet. using knife in chef i can spawn a new rackspace instance, install and configure my entire stack, and deploy my application all in one command. i can launch an exact copy of an application in 5 - 10 minutes - during which i do nothing but sit and watch the entire chef-client run go by. it's about as turn-key as it gets. of course writing out our entire configuration process in ruby is a bit of work - but it's front-loaded work that you never have to repeat and saves you months of tweaking and copying configuration files. you can subsequently modify quite easily to fit new deployments and new applications.
if you need it to be quicker you simply take a snapshot of the resulting instance and clone it to new instances.
>Why is it so hard to have one-click templates that give you a server that can run 90% of what a user needs,
well, the thing is? yes, hardware costs are falling (and sysadmin costs are rising) to the point where it's actually cheaper for me to give you a VPS than to give you a shared hosting account.
Problem is... that's because I'm depending on you to do the sysadmin work. That's my problem with things like Cpanel and webmin, well, and pre-installed services, too. they give folks a false sense of control. I mean, with control panels, I say they work 95% of the time. The problem is that then you expect it to work all the time; you depend on it. so 5% of the time? they are kinda fucked. And if you want to take that risk, that's fine, sometimes that's rational. but I don't want to be the guy who has to clean up that 5% of the time... not for $10 a month or whatever you are paying, that's for sure.
That's why I don't like the idea of pre-installed applications in an image... As a service provider? I don't want to set the expectation that I'm going to support apache/php/ruby/node/mysql/postgresql. and pre-installing it does set the expectation, for a lot of people, that you will support it when it breaks six months from now.
give you a server that can run 90% of what a user needs
Are you saying that you just keep adding things to the server until you've covered that mythical user?
That sounds like an absolute disaster from a security, maintenance, and resource usage perspective.
I don't see what problem this solves. Start an EC2 instance (takes seconds), configure it as you want it (which to me includes a hardened, specific set of software with very specific routing rules where nginx sits in front of node, etc), and then save an AMI from your specific mix of technology.
Trivial, and perfectly suited to the user's needs. I can then spin up any number of perfectly suited instances.
>That sounds like an absolute disaster from a security, maintenance, and resource usage perspective.
Installed and available are a huge sight different from running and exploitable, though. I can have a completely unpatched Apache from 5 years ago, it's invulnerable to foes up to and including the NSA unless the app is running and listening on a port :)
Would be trickier, but not impossible for more system-level apps like Cpanel/Virtualmin.
In the context of web applications, would it make sense to stop using the word "app" to refer to software/daemons available in the distribution repositories, and instead only use "app" to mean the products and services built on the former?
