Can anyone speak to who this sort of things is targeted at? The way I see it:
If uptime is really important, you’re going to need your app running across multiple machines anyways, in which case doing zero downtime deployments is a pretty standard pattern of shifting traffic to one machine, upgrading the other, and then letting traffic come back.
If uptime is not so important to your application, then 10 to 20 seconds of downtime doesn’t seem like a big loss. You could say “hey it’s just a better experience” which is roughly what the article says but I struggle squaring that with the running on one machine thing.
I agree, but the technique described in the article is similar in a multiple machines setup. Instead of having 2 machines, they have 2 processes (the old one and the new one) running on the same machine, with nginx acting as a load balancer.
Can anyone speak to who this sort of things is targeted at? The way I see it:
I agree, but the technique described in the article is similar in a multiple machines setup. Instead of having 2 machines, they have 2 processes (the old one and the new one) running on the same machine, with nginx acting as a load balancer.
Interesting. The links to the Gists don’t work.