Swapping order of half_half strategy to ensure maximum uptime

Cimpress-MCP / vagrant-orchestrate

Cross platform deployment orchestration to existing managed servers using vagrant provisioners

Other

19 stars 13 forks source link

Swapping order of half_half strategy to ensure maximum uptime #53

Closed briangweber closed 7 years ago

briangweber commented 7 years ago

Use case: we have 2 machines running a service, we want to add 2 more. We create the instances and then re-run the deployment task to push to the added instances. Instances 1 and 2 are deployed first, but since they are the only ones actually running the service, the deployment can cause a brief outage while the service restarts. If we deploy to the back half first, we can avoid this problem.

briangweber commented 7 years ago

I ran into this same problem testing locally - was hoping it was just due to my environment.

chrisbaldauf commented 7 years ago

Code change looks good to me but it appears that the CI build failed due to some dependency mismatches.

Alternatively, could you intersperse your new instances into your vagrantfile when deploying to new boxes?

Start with

old_instance_1
old_instance_2

And then transform it into

old_instance_1
new_instance_1
old_instance_2
new_instance_2

briangweber commented 7 years ago

I'm honestly unlikely to invest any more time in this since we're moving to docker for deployment shortly. The code fix only took a couple minutes, but I suspect fixing the build is going to take awhile.