This documentation is for Deis Workflow (v1). For v2 documentation visit https://github.com/deis/workflow.
Deis uses a service oriented architecture with Components grouped into a Control Plane, Data Plane and Router Mesh.
Operators use the
Deisctl API to stand up the cluster’s Control Plane, Data Plane and Router Mesh.
End-users of the platform interact with the Control Plane using the
The Control Plane dispatches work to the Data Plane via a scheduler. The Router Mesh is used to route traffic to both the Control Plane and Data Plane. Because the router mesh is usually connected to the public Internet, it is often connected to a front-end load balancer.
The Control Plane performs management functions for the platform. Control plane components (in blue) are all implemented as Docker containers.
The Store component consists of a number of smaller components that represent a containerized Ceph cluster which provides a blob storage API and POSIX filesystem API for the control plane’s stateful components:
The Data Plane is where Containers (in blue) are run on behalf of end-users.
The platform scheduler is in charge of placing containers on hosts in the data plane. Deis also requires a few lightweight components on these hosts:
The Router Mesh publishes Applications to consumers.
Each Router in the mesh is a configurable software load balancer designed to expose
Containers running in the data plane.
Routers track healthy containers using a distributed, watchable store like
Any changes to router configuration or certificates are applied within seconds.
For small deployments you can run the entire platform – Control Plane, Data Plane and Router Mesh – on just 3 servers.
For larger deployments, you’ll want to isolate the Control Plane and Router Mesh, then scale your Data Plane out to as many servers as you need.
See Isolating the Planes for further details.
The Deis Control Plane, Data Plane, and Router Mesh components all depend on an etcd cluster for service discovery and configuration. For larger deployments, you’ll want to isolate this as well.
Those building larger clusters should reference Isolating etcd for further details.