My ASP.Net site runs as farm of Windows EC2 web servers. Due to a recent traffic surge, I switched to Spot instances to control costs. Spot instances are created from an AMI when hourly rates are below a set rate. The web servers do not store any data, so creating and terminating them on the fly is not an issue. So far the website has been running fine.
The problem is deploying updates. The application is updated most days.
Before the switch to a Spot fleet, updates were deployed as follows (1) a CI server would build and deploy the site to a staging server (2) I would do a staggered deployment to a web farm using a simple xcopy of files to mapped drives.
After switching to Spot instances, the process is: (1) {no change} (2) deploy the update to one of the spot instances (3) create a new AMI from that deployment (4) request a new Spot fleet using the new AMI (5) terminate the old Spot fleet. (The AMI used for a Spot request cannot be changed.)
Is there a way to simplify this process by enabling the nodes to either self-configure or use a shared drive (as Microsoft Azure does)? The site is running the Umbraco CMS, which supports multiple instances from the physical location, but I ran into security errors trying to run a .Net application from a network share.
Bonus question: how can I auto-add new Spot instances to the load balancer? Presumably if there was a script which fetched the latest version of the application, it could add the instance to the load balancer when it is done.
