thoughts_on_stackmachine.md

Ok, so here's my laundry list. I'm going to use Chef terminology here, but this bleeds over to Docker, Salt, Puppet, etc.

Your stack creation product should force me to use best practices. I should specify the project type and you should do the following.

• prompt me for my source code repo
• set up apache/nginx/caching/db
• run my project

It'd be great for this to work out of the box for major frameworks like rails, django, pyramid, express

Along with that, the folliowing should be included.

#All Projects

• separation of configuration and source code
• nfs mount for editing code on the host machine
• logrotate
• splunk || splunk storm || loggly integration
• sane default CloudWatch metrics like memory usage
• Autoscale friendly deployments (mounted ebs code dir, ebs snapshot on deployment, autoscale user_data pulls from latest snapshot)
• deployments that don't depend on github (tarball pushed to s3 on each commit & installed from on deploy)

#Rails

• passenger in dev and production to tease out threadsafe problems for libs like Redis.
• sane passenger defaults for MaxPoolSize and friends for development and production, based on instance type chosen.

#Python

• virtualenvs that build on each deployment from requirements.txt or pip stuff. Don't let system site_packages leak through to developoment
• supervisor configuration

#DBs

• RDS support OR
• EBS snapshotting using ec2-consistent-snapshot
• Raid Support
• mysql dumps

#Vagrant

• multi-node deployment mode (several boxes)
• single-node deployment same boxes, different ports