Skip to content

Latest commit

 

History

History
40 lines (32 loc) · 2.39 KB

release_plan.md

File metadata and controls

40 lines (32 loc) · 2.39 KB

8/30 Release

Features

  1. Edge clusters
  2. Cascading edge clusters
  3. Inter-cluster communication

Edge Cluster

The goal of this feature is to allow running K8s-flavored cluster on the edge and can both tolerate network and node failure. In specific:

  • K8s cluster running in the cloud as the cloud control plane, while various K8s flavored clusters (e.g. vanilla K8s, Arktos, etc.) running on the edge
    • Status of edge clusters and edge workloads can be obtained through the cloud control plane via kubectl
    • Workloads (e.g. deployments, jobs, etc.) run in edge clusters, and workload CRUD can be managed from the cloud
    • Workloads (Deployment, pod) created from cloud control plane targeting edge clusters are propagated to the edge following filters
  • Failure tolerance
    • Edge clusters and workloads continue functioning when the network with cloud disconnects
    • Edge clusters and workloads resume status and workload syncing when network connection recovers
    • Edge workloads continue functioning if resource allows when edge cluster nodes fail (K8s behavior)

Cascading Edge Clusters

The goal of this feature is to allow multiple edge clusters to be cascaded in hierarchical form. In specific:

  • K8s flavored clusters (vanilla K8s, Arktos leaf cluster) running on the edge can be attached to another edge cluster
  • Cluster and workload status are propagated to the cloud control plane in two modes
    • Regular heart beat (longer interval, Mission & edge cluster)
    • Change-based reporting (e.g. every 10s)
  • Edge clusters can be specified by filters representing geological or organizational topology

Inter-Cluster Communication

The goal of this feature is to allow workloads (pods) to communicate with each other when distrubted into multiple edge clusters. In specific:

  • A gateway for a edge cluster to establish a communication "channel" with other clusters on the edge
  • Network workloads (e.g. VPC, subnet) can be created in selected edge clusters
  • (Mizar) Applicatio workloads such as pods can be created into the specified VPC and subnet
  • (Mizar) Applicatio workloads (pods) in the same edge cluster and VPC could communicate
    • Both pods in the same subnet
    • Pods in different subnets
  • Applicatio workloads (pods) in different edge clusters but same the same VPC could communicate
    • Pods in different subnets