You already have alerts and images (snapshots/backups), now all you have to do, is setup some config that when XYZ alert goes off (80% CPU) auto spin up a new droplet (size determined by config) and attach that to the load balancer. This Droplet would start accepting traffic once the health check passes.
As a phase 2, when certain alerts get hit (cpu below 20% for 60 minutes), then de-scale by killing off the youngest droplet.
It seems like all the pieces of the puzzle are there, and it appears that nobody has asked for this specifically like this before on here. Please correct me if I'm wrong.
This is absolutely needed before we can truly consider running a production site on DO.
Failed servers should be replaced by similar servers (saved images) (MUST HAVE)
If certain load metrics are exceeded, additional servers should be started (REALLY NICE TO HAVE)
The opposite of (2) in order to scale down.
Attachments Open full size
Not just autoscaling, but killing droplets that are malfunctioning and replacing them!! Right now if a Droplet is failing the Load Balancer just stops sending traffic to it and I think we get an alert, but I might be wrong on that since I am just now using DOs load balancer for the first time on a new project.
Â
AWS load balancer is so much nicer because it can autoscale up and down, and destroy an instance if it fails and replace it. Unfortunately aws is too expensive for startups, and people like me prefer DigitalOcean over AWS or Google.
Attachments Open full size
Minor point: if you killed off the oldest droplet, it would help with configuration drift and other bugs which only manifest in long-running systems.
Attachments Open full size
This would be really nice for my projects, I could see me using this
Attachments Open full size