Future of bosh-lite: Docker, Otto, multiple Vagrant providers?

[SamuelMarks]

Discussion thread
Bosh-lite is a great project for getting a development environment up and running moderately quickly.

Had to buy another 8GB of RAM for it, and then it only took ~90 minutes to install everything (cf v230 and diego). The VM allocated 6144MB of memory for itself, and 4 CPUs.

In production deployments there's a multitude of specialised nodes.

Why don't we make our dev install micro-service orientated?

Options
Here are the primary options I can think of, add yours if its not below:

• An 'ocean' of Dockerfiles, orchestrated trivially (we're talking one host afterall)
• Otto (Hashicorp's Vagrant successor; uses Vagrant beneath-the-scenes)
• Vagrant providers (includes Docker)

Disadvantages

• Different to current approach
• Development required

Advantages

• Easier to isolate and test individually, and progressively test with more dependent components
• Less system overhead
• Smaller file-size (great for sharing 'snapshots' across your development teams)

[joek]

I am not sure if the intend from bosh-lite is to setup your development environment. It is from my prespective more something to use, if you like to test bosh releases and depoyments on your local box (or on AWS, using the AWS Vagrant provider). The benefits are, the deployment is running a bit faster and it is using less resources.

[SamuelMarks]

Perhaps. Either way, my proposal that a new approach to bosh-lite is required is still to be addressed.

IMHO the Docker solution is probably the nicest, microservice all 12+ cf VMs and that should be useful for production usage [eventually] also. Obviously with some sort of overarching fs such as Ceph, Gluster or abstracted object storage (linking into S3 et al.).