Question, How are you cutting over your application instances when doing updates?

Hi there

I wanted to pose a question to the community to see if anyone would be willing to share for how they are handling production application updates? Specifically, when the application is an api (or website) that has a high volume of concurrent traffic.

I’m keen to understand if your deployment workflow has been incorporated into the nomad.job or it’s a workflow coordinated externally from nomad, ie you might canary deploy, reconfigure load balancer, then promote your app in nomad?

For reference I am trying to design a deployment workflow in preparation for moving the api to nomad for our production traffic, so I’m currently simulating the load to test. My lab is using Fabio (also hosted by nomad) to load balance our api instances. I am not tried to Fabio. I currently canary deploy, use a minimum wait and then auto promote but during that cut over I’m see a few requests fail from my client which I would like to avoid.

I appreciate any advice people would be willing to share.