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
Before stopping developing Garden, we were plans to delegate more and more work to containerd, and to eventually get rid of our runc backend. We should document these plans, and their blockers.
We couldn't find a way to implement the Volumizer interface using snapshotter, it felt like container creation and rootfs creation were tightly coupled.
We considered creating a new Gardener (that is, a new implementation of garden.Backend), fully backed by containerd.
This is probably impossible without refactoring the way peas are implemented. We want the peas flow to create containers using Gardener instead of depending directly on low-level implementation details.
Before stopping developing Garden, we were plans to delegate more and more work to
containerd
, and to eventually get rid of ourrunc
backend. We should document these plans, and their blockers.containerd
without processes modegrootfs
with thesnapshotter
in fullcontainerd
modeVolumizer
interface usingsnapshotter
, it felt like container creation and rootfs creation were tightly coupled.Gardener
(that is, a new implementation ofgarden.Backend
), fully backed bycontainerd
.Gardener
instead of depending directly on low-level implementation details.runc
modecontainerd
due to issues with the shim freezing (see Garden failing with error "open mnt namespace: No such file or directory" #154 and #176654373 on Tracker).The text was updated successfully, but these errors were encountered: