mustwin / cmeter Goto Github PK

cAdvisor will need to be embedded for us to abstract the nuances of the various container runtimes and make metric collection easier.

Kubernetes's Kubelet similarly embeds cAdvisor and would be a good starting point:
https://github.com/kubernetes/kubernetes/tree/master/pkg/kubelet/cadvisor

there's a good use case for a reporting driver that publishes events to kafka

This is mutually exclusive to the agent mode and should spin up an HTTP server that logs the events entities it receives at the POST /events endpoint and responds with status code 204.

start collection at the configured sampling rate when a container enters or is in the running state.

listen for container creation events from cadvisor and publish a running state change event to the API

Requires MustWin/cToll#12 to be done. Use the github.com/MustWin/ctoll/ctoll/api/client package to send events.

make reporting destination a generic interface

The agent command should start the primary run mode of the application:

$ cmeter agent

identify any active trackable containers on startup and register them locally in the running state.

create a mock reporting driver

The api command should start the mock API server for development and testing.

$ cmeter api

create a containers implementation to interface with external cAdvisor nodes

This will let the reporting be better tailored to each driver's scenario.

listen for container deletion events from cadvisor and publish a stopped state change event to the API

add config value

convert the api client stuff into a reporter implementation

log out events received just remove this for now

just POST the event to the server url (specified by parameter) as-is, no manipulation. Include cmeter version and configurable headers

only track metrics and state events for containers that are identified with the targeting label.