You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository has been archived by the owner on Feb 24, 2020. It is now read-only.
If image example.com/a depends on example.com/b and if, for any reason, example.com/b changes (updated ACI on the remote server, new latest image url discovered, fetch from file etc...) the rendered image is different.
This isn't a problem with rkt without overlayfs as the image is rendered at any run/prepare but it's a problem with the treestore as it's used as the lowerdir of the overlayfs for that image.
Actually, the treestore id is the upper image hash. Since, in this case, the upper image hash isn't changed, it cannot be simply overwritten because it can be in use by a pod's overlayfs
I see two main needs:
Find a different way to name the treestore rendered images so they can be multiple for the same upper image and correctly referenced and cleaned.
Detect at rkt prepare/run time when a treestore should be re-rendered.
The text was updated successfully, but these errors were encountered:
If image
example.com/a
depends onexample.com/b
and if, for any reason, example.com/b changes (updated ACI on the remote server, newlatest
image url discovered, fetch from file etc...) the rendered image is different.This isn't a problem with rkt without overlayfs as the image is rendered at any run/prepare but it's a problem with the treestore as it's used as the lowerdir of the overlayfs for that image.
Actually, the treestore id is the upper image hash. Since, in this case, the upper image hash isn't changed, it cannot be simply overwritten because it can be in use by a pod's overlayfs
I see two main needs:
rkt prepare/run
time when a treestore should be re-rendered.The text was updated successfully, but these errors were encountered: