roadmap.md

Kubespray's roadmap

Kubeadm

• Propose kubeadm as an option in order to setup the kubernetes cluster. That would probably improve deployment speed and certs management #553

Self deployment (pull-mode) #320

• the playbook would install and configure docker/rkt and the etcd cluster
• the following data would be inserted into etcd: certs,tokens,users,inventory,group_vars.
• a "kubespray" container would be deployed (kubespray-cli, ansible-playbook, kpm)
• to be discussed, a way to provide the inventory
• self deployment of the node from inside a container #321

Provisionning and cloud providers

• Terraform to provision instances on GCE, AWS, Openstack, Digital Ocean, Azure
• On AWS autoscaling, multi AZ
• On Azure autoscaling, create loadbalancer #297
• On GCE be able to create a loadbalancer automatically (IAM ?) #280
• TLS boostrap support for kubelet #234 (related issues: kubernetes/kubernetes#20439
  kubernetes/kubernetes#18112)

Tests

• Run kubernetes e2e tests
• migrate to jenkins (a test is currently a deployment on a 3 node cluste, testing k8s api, ping between 2 pods)
• Full tests on GCE per day (All OS's, all network plugins)
• trigger a single test per pull request
• single test with the Ansible version n-1 per day
• Test idempotency on on single OS but for all network plugins/container engines
• single test on AWS per day
• test different achitectures : - 3 instances, 3 are members of the etcd cluster, 2 of them acting as master and node, 1 as node - 5 instances, 3 are etcd and nodes, 2 are masters only - 7 instances, 3 etcd only, 2 masters, 2 nodes
• test scale up cluster: +1 etcd, +1 master, +1 node

Lifecycle

• Adopt the kubeadm tool by delegating CM tasks it is capable to accomplish well #553
• Drain worker node when upgrading k8s components in a worker node. #154
• Drain worker node when shutting down/deleting an instance
• Upgrade granularity: select components to upgrade and skip others

Networking

• romana.io support #160
• Configure network policy for Calico. #159
• Opencontrail
• Canal
• Cloud Provider native networking (instead of our network plugins)

High availability

• (to be discussed) option to set a loadbalancer for the apiservers like ucarp/packemaker/keepalived While waiting for the issue kubernetes/kubernetes#18174 to be fixed.

Kubespray-cli

• Delete instances
• kubespray vagrant to setup a test cluster locally
• kubespray azure for Microsoft Azure support
• switch to Terraform instead of Ansible for provisionning
• update $HOME/.kube/config when a cluster is deployed. Optionally switch to this context

Kubespray API

• Perform all actions through an API
• Store inventories / configurations of mulltiple clusters
• make sure that state of cluster is completely saved in no more than one config file beyond hosts inventory

Addons (with kpm)

Include optionals deployments to init the cluster:

Monitoring

• Heapster / Grafana ....
• Prometheus

Others

Dashboards:

• kubernetes-dashboard
• Fabric8
• Tectonic
• Cockpit

Paas like

• Openshift Origin
• Openstack
• Deis Workflow

Others

• remove nodes (adding is already supported)
• being able to choose any k8s version (almost done)
• rkt support #59
• Review documentation (split in categories)
• consul -> if officialy supported by k8s
• flex volumes options (e.g. torrus support) #312
• Clusters federation option (aka ubernetes) #329