Ask HN: Reproducable deployment Workflow for Linode?
we're building a machine-learning powered analytics platform running on Linode. Have you found a good workflow to manage your services on your Linode VPNs? Willing to share with the community? ️
Requirements on our vanilla Linode instance:
- Deploy/Share “Security-Profiles” to VPNs - Fast Backups/Snapshots & Rollbacks (ZFS/Btrfs or something else?) - Every Step is Reproducible - Replayable (SSH-)Sessions for Training - Simple, but Scalable Architecture - Just “Build -> Test -> Ship -> Run” - Eventually have Monitoring & Failover
There is Vagrant-Linode[1] and I intend to keep services isolated using Docker. Does using vagrant+docker even make sense? How else can I put the requirements above into a workflow? What do you recommend instead?
Let's assume we have a vanilla Linode instance.
• How'd you make any changes on the server (reproduably)?
• Is Debian the only OS offering reproducable builds?
• Should we use Kubernetes, Mesos whatnot?
• What about Stackscripts, or would you skip that and take Ansible for configuration management instead?
- [1] https://github.com/displague/vagrant-linode/