controller

README

Last.Backend OpenSource

API server

Controller

Meta:

TODO Describe runtime workflow

Workflow:

Controller workflow:

Cluster observer

• TODO: Cluster observer runtime: [ ] TODO: Cluster observer struct with methods: - [ ] TODO: Restore cluster state: - TODO: [x] start cluster IPAM - TODO: [x] get cluster information - TODO: [x] get all nodes in cluster - TODO: [ ] for each node - run node observer - TODO: [ ] sync current cluster state - TODO: [ ] get network information - TODO: [ ] get all endpoints - TODO: [ ] get all cluster manifests - [ ] Runtime: - [ ] Watch node stats changes - [ ] Watch requests for node lease - [ ] Watch requests for node release - [ ] Lease node: - [ ] Find node matches scheduling type - [ ] Calculate new cluster state - [ ] Return node - [ ] Release node: - [ ] Release node state - [ ] Calculate new cluster state - [ ] Endpoint: - [ ] Provision: check endpoint spec and network spec - [ ] create new endpoint and save it to state storage - [ ] update endpoint spec if previous condition is true - [ ] remove endpoint if network spec is empty

[ ] Node observer: [ ] Node observer struct [ ] Restore: restore current node state: - [ ] get all manifests for node - [ ] calculate memory usage - [ ] calculate storage usage - [ ] check network is ready - [ ] calculate node global state: Node state conditions: - CNI state: [Ready, Error] - [ Warning ] state - CPI state: [Ready, Error] - [ Warning ] state - CRI state: [Ready, Error] - [ Not Ready ] state - CSI state: [Ready, Error] - [ Warning ] state [ ] Update: update current node state [ ] Remove: clean up node and remove [ ] Manifest: add manifest to node [ ] Manifest: set manifest to node [ ] Manifest: del manifest from node

[ ] Namespaces observer

[ ] Namespace observer [ ] Namespace observer struct [ ] Restore namespaces observer: - [ ] Get all namespaces - [ ] Loop over namespaces and get services for each - [ ] Loop over services and create service controller - [ ] Start service controller loop [ ] Create watcher for service - [ ] if service state is initialized -> pass service for update - [ ] if service is not initialized -> initialize [ ] Create watcher for pods - [ ] if service for this pod is initialized > pass pod for service update - [ ] if service is not initialized > skip [ ] Runtime loop: - [ ] Start service watching changes

[ ] Services observer Service observer should watch service state and manage service runtime Restore: service controller state - [x] get all pods - [x] get all deployments - [x] get endpoint for service - [x] loop over pods and pass pod to service update - [x] loop over deployment and pass deployment to service update - [x] check current service spec and provision service if needed Handle changes: Runtime loop: Scale: Update: Remove:

[ ] Pod package: Pod status states: created > provision > ready > destroy > destroyed > error

[x] Create: Create new pod [x] Destroy: Mark Pod for destroy [x] Provision: Provision pod [x] Remove: Remove pod if previous state was destroy

[x] Deployment package: Deployment status states: crated > provision > ready > destroy > destroyed > error

[x] Create: Create new deployment based on service spec [x] Scale: Scale deployment for replicas, it should work as increment pods and decrement [x] Destroy: Mark all pods for destroy [x] Remove: Remove deployment if service is marked as removed [x] Cancel: Cancel deployment if not need anymore

[x] Service package: [x] Remove: Remove service [x] Sync: Service sync state [x] Provision: Provision service [x] Destroy: destroy service

[x] Endpoint package: [x] Create: create new endpoint [x] Update: update endpoint spec [x] Destroy: destroy endpoint

Documentation

Index

• func Daemon() bool

Functions

func Daemon

func Daemon() bool

Daemon - controller entrypoint