How to rolling restart the old container after the new container be healthy in case of the downtime?

[dusty-cjh]

Currently, between shutting down the old container and starting the new container, there is several seconds downtime. Since the service need to connect with MySQL / Redis, read config file or etc.
My question is how to roll restart the old container after the new container is healthy?

I have tested the --rolling-restart option, and I think that is useless, you can see obvious HTTP failures below(about 15%):

So is there have any solutions?

My scenario is to deploy new container on multiple machine without introducing k8s (to temporarily mantain compat with the old procedure).

Since this should be a very common requirement, but I don't find any existing solution(exclude swarm and k8s), that's weird.

Could you give me some advice? Thanks a lot!