cloudify-incubator / cloudify-kubernetes-provider Goto Github PK

Update this blueprint to allow option to install and configure Provider/Autoscaler.

Regarding this

Currently, we resolve instances based on node id. I think type would be better. The problem is we don't have a type yet.

We could resolve based on a node type like this, or, since in aria, such a type does not exist, we could resolve based on a runtime property like "RESOURCE_TYPE".

For now delete workflow ignore filed with hits for node names for delete, need to change code to delete nodes only defined as nodes for possible delete. Use scalelist{up/down} for scaling.

Need to change kubernetes host nodes and load balancers nodes.

Python is better for working with Cloudify blueprints because it is awkward to use the context in Bash, and YAML/JSON/dict manipulation in general is awkward in Bash.

We have this script for releasing our plugins.

We need to do the same thing for tagging a release in the provider and publishing binaries.

Update the current load balancer behavior.

Today, when a user "expose" a pod or deployment using a load balancer method, the provider returns 127.0.0.1. We need to provide the real URL of the node instance. It should default to public_ip runtime property, and fall-back to ip runtime property.

For now we hardcoded cloudify manager ip by{ get_attribute: [ kubemanager, cloudify_agent, broker_ip ] }

Need to use loadbalancer or get ip information from agent config on node.

Kubernetes Nodes should be labelled with the deployment that created them (optionally off).

If I have already created a cluster in some IaaS, Azure for example, and I want to add a VM to the cluster with a specific pod or service, I need a infrastructure blueprint that I can import.

Please create such a blueprint for:

• AWS
• Azure
• GCP
• Openstack
• Vsphere

• The blueprint should enable labeling nodes as well.

• Current implementation return instance after instance receive hostname, we need to wait while all executions related related to instances will be finished.
• Before delete any node need to wait several minutes before really delete node, so stop delete any nodes younger than 15 minutes.
• Implement CloudifyScaleProvider::Refresh in autoscale repo for cache Scaling Group/Instances hierarchy.

• Using libvirt plugin
• Using host pool plugin

Update cfy-go client installation on provider to use SSL.
Also the build/cluster creation blueprint should configure generate the scripts as well.

Currently Instances are mapped to nodes in the cluster blueprint of node type "kubernetes_host".Ref

The problem with this approach is that the node type needs to be redefined for each cluster deployment, examples:

• cloudify-kubernetes-provider/examples/cluster_blueprint/aws.yaml

  Line 34 in 56017e1

  kubernetes_host:

• cloudify-kubernetes-provider/examples/cluster_blueprint/azure.yaml

  Line 48 in 56017e1

  kubernetes_host:

On the one hand, since a Kubernetes Node is strictly speaking a machine, the direct corresponding type in cloudify is cloudify.nodes.Compute`. On the other hand, a Kubernetes Node should also have several capabilities that are not part of Cloudify Compute nodes - container runtime, kube-proxy, etc.

I want to propose an alternative:

A new node type that is "contained_in" a Cloudify Compute node that represents the Kubernetes Node, such as:

node_types:
    cloudify.nodes.Kubernetes.Node:
        derived_from: cloudify.nodes.SoftwareComponent
        properties: {}
node_templates:
    kube_node:
        type: cloudify.nodes.Kubernetes.Node
        relationships:
        - type: cloudify.relationships.contained_in
           target: some_vm

This feature include the following changes/PRs:

• 1. #49
• 2. cloudify-incubator/kubernetes-node-blueprints#1
• 3. cloudify-incubator/kubernetes-lb-blueprints#1
• 4. cloudify-cosmo/cloudify-gcp-plugin#84
• 5. cloudify-incubator/cloudify-rest-go-client#12
• 6. cloudify-incubator/kubernetes#1
• 7. https://github.com/cloudify-incubator/kubernetes-autoscaler/pull/1/files
• 8. cloudify-examples/simple-kubernetes-blueprint#6

@EarthmanT @0lvin Please take a look to the following PRs