3.2 KiB
containerd roadmap
This is a high level roadmap for the project that outlines what is currently being worked on, what comes next, and where you can help.
The following are the different status the various phases of development can be in:
- Not Started - no work or thinking has been done towards the goal
- In Design - design work has started for the component and you can find design documents in the
designfolder - In Progress - design has mostly finished and development has started
- Completed - the development work has been completed
- Stable - the apis for the phase are feature complete and considered stable
We would like to follow the roadmap and develop the components one by one to completion before starting the next phase. If PRs are opened for another phase before the previous phase has been completed they will be closed as we are not ready for them at that time.
Phase 1
Status: In Progress
GRPC API
Documents:
We are going from a top down design for filling out this missing pieces of containerd and design of the API.
Design
Documents:
The high level design work is needed so that the architecture of containerd stays consistent throughout the development process.
Build & Test Process
Documents:
We need to have a simple build and test process for new developers to bootstrap their environments. Because containerd will be the base of many high level systems we need to have a simple build process that does not require high level tooling.
Phase 2
Phase 2 includes most of the design and development work for the execution and storage layers of containerd. It will include porting over existing "graph drivers" from Docker Engine and finding a common model for representing snapshots for layered filesystems.
This will also include moving the existing execution code support OCI's Runtime Spec and the existing containerd execution code.
Status: In Design
Runtime
The runtime layer is responsible for the create of containers and the management and supervision of processes.
Storage
Documents: https://github.com/docker/containerkit/blob/master/design/snapshots.md
The current graph drivers were built when we only had overlay filesystems like aufs. We forced the model to be designed around overlay filesystems and this introduced a lot of complexity for snapshotting graph drivers like btrfs and devicemapper thin-p. Our current approach is to model our storage layer after snapshotting drivers instead of overlay drivers as we can get the same results and its cleaner and more robust to have an overlay filesytem model snapshots than it is to have a snapshot filesystem model overlay filesystems.
Phase 3
Phase 3 involves porting the network drivers from libnetwork and finding a good middle ground between the abstractions of libnetwork and the CNI spec.
This also includes getting support for the OCI Image spec built into containerd.
Status: Not Started
Distribution
Networking
The networking component will allow the management of network namespaces and interface creation and attachment to namespaces.
Phase 4
Phase 4 includes work on helping with the releases and packaging of containerd for various distros.
Status: Not Started