Allow Droplets to be automatically bootstrapped with Chef
It would be *sick* if it were possible to have systems be automatically bootstrapped with Chef upon creation. e.g. Users could set up recipes and roles in the DO control panel somehow, and then pick a role to automatically create a Droplet from.
Additionally, users could share their Chef stuff with each other, which would allow everyone to benefit from the magical auto-provisioning. And, it's more transparent than allowing users to share pre-built images, since anyone can inspect the code for the Chef recipes before using them to provision a Droplet.
Or, even better than managing Chef stuff through the DO control panel, it would be great if I could just give a URL to a GitHub repo containing the recipes/roles, and then the Droplet would automatically pull down this repo and bootstrap itself with the desired role(s).
Cody Krieger
shared this idea
14 comments
-
Vincent Gijsen
commented
The way we use Chef, is from our command line, not a web-gui.
We've created a simple script which starts a new node using the Digital-ocean API, and uppon node creation, we query the new node's ip-adress, issue the appropriate knife command to bootstrap it, and voila.
no mouse clicking involved there.
imho spending time to make this work using a gui, isn't very useful, as most users of chef will typically deploy there own scripts (or variants)
-
Anonymous
commented
Chef is preferable for myself. Currently using AWS opswork for deploying chef based stacks, would be nice to port some of that stuff to DigitalOcean ;)
-
Tomás Gutiérrez
commented
Hi all - the beauty of these services is that you can make them agnostic to the API. A friend just told me about DO, and I found this:
https://github.com/rmoriz/knife-digital_ocean
Similarly to AWS/RS, you can use knife to bootstrap a droplet and fulfill your DevOps happy bucket.
Hit me up @togume if you have any questions.
Have fun, and looking forward to playing with DO.
-
James Seigel
commented
To fill out the list, look up sprinkle.
-
Fabric is cool but it's been a while since I've heard someone mention it, is it still being widely and actively used or has it been replaced more by puppet/chef?
-
Micheal Cottingham
commented
If you guys have an infrastructure for Salt (sweet!) or Chef or Puppet so I don't have to build it out, I'd happily pay an extra nominal fee to use it. I started with Fabric because I really don't want to have a master/slave setup with a Puppet or Chef or Salt exposed that can get popped. That's a perfect pivot in to the rest of my architecture ... For now, I'm learning Fabric as it is not a pub/sub but a push arch.
-
Hopefully 2-3 months which will overlap either with pre-built images or just stacks that are built atop of chef or puppet for easy roll-out.
-
Jonathan
commented
Hi Moisey, any idea on an ETA for this? (weeks, months, etc)
-
Integration with Salt is also coming =]
-
Kenn Ejima commented
IMO chef (or puppet) is way too complicated for average VPS users.
Thus I built Sunzi, which is a super lightweight alternative - it even has official Linode support to setup / teardown VMs from CLI. I'm going to add DigitalOcean support when the API is fully implemented.
-
Roland Moriz
commented
You can use my knife plugin to create a droplet and bootstrap it with chef:
https://github.com/rmoriz/knife-digital_ocean(one command with chef-server provisioning, using "knife-solo" it's two commands)
-
We're looking into it but definitely months away at this point.
More likely to be puppet based if it does get rolled out.
-
Ankur Sethi
commented
If a user is into this, they can create their own image and boot from it. Installing software on base images means they would have to be separate Chef templates.
-
We currently do not have this in the pipeline but we do know of a couple of startups that are operating in this space.
Basically managed Chef/Puppet where you can spin up servers and they will get auto-signed and have that deployed.
So this may be something that comes together a bit later when we've cleared through a bit more of the infrastructure backlog
