Bogdan Dobrelya 72a25bfd06 Update roadmap for the kubeadm LCM track

Signed-off-by: Bogdan Dobrelya <bdobrelia@mirantis.com>

2016-10-18 13:44:45 +02:00

Kargo's roadmap

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 (kargo-cli, ansible-playbook, kpm)
to be discussed, a way to provide the inventory
self deployment of the node from inside a container #321

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: https://github.com/kubernetes/kubernetes/pull/20439
https://github.com/kubernetes/kubernetes/issues/18112)

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

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

### 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.

Delete instances
kargo vagrant to setup a test cluster locally
kargo 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

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

Include optionals deployments to init the cluster:

##### Dashboards: