We pull BaseImages every time it's missing from the docker host.
We build and push Imagess every time it's missing from the docker host.
This is not needed and causes a surge of activity that hogs cpu/mem/network whenever you kill an image-manager pod and a new one is started.
New Behavior
PullBaseImages/'Imagess only if it's missing and a certain time has elapsed since last successful pull.
We will stop building and pushing Image if the image is just in missing state. Instead, a pull would suffice to verify that the container image exists on the registry.
TODO
Prune images/base-images from docker host based on some cache eviction strategy. This will ensure that we don't eventually pull all images/base-images on the docker host.
Current Behavior
BaseImage
s every time it's missing from the docker host.Images
s every time it's missing from the docker host.This is not needed and causes a surge of activity that hogs cpu/mem/network whenever you kill an image-manager pod and a new one is started.
New Behavior
BaseImage
s/'Images
s only if it's missing and a certain time has elapsed since last successful pull.Image
if the image is just in missing state. Instead, a pull would suffice to verify that the container image exists on the registry.TODO Prune images/base-images from docker host based on some cache eviction strategy. This will ensure that we don't eventually pull all images/base-images on the docker host.