colearendt
commented
1 year ago Howdy @pat-s ! So glad to hear things are working well for you!
Unfortunately, there is not a way around this today 😞 We do have some work scheduled for longer term letting the Connect hosts "collectively" manage a collection of pods, so min_processes = 3 could be honored in the way you expect.
In the meantime, I would recommend setting min_processes = <desired> / 3 so you would set min_processes = 1 and understand that each node will honor that minimum, so you get 3 pods. This is consistent with how minimum processes worked in the traditional execution environment (min processes on each node). So we have a bit of work to do to refactor that setup!
I will pass this feedback along!!
Hi @colearendt!
We're happily running Connect in HA with 3 pods. Works great WRT to rolling updates and HA in general.
However, each connect app also gets replicated X times as the respective job is executed by each server replica. Is there a way to adjust this? This behavior becomes quite expensive for production apps that take let's say 500 MB mem and
min_processes = 3- which results in > 10 pods reserving 500 MB RAM.I know one option could be to reduce the overall replicas but I would really like to stay with HA for the server pod.