Cluster-svc service

cluster-svc runs as a notifyee of MEX controller and listens to cluster instance notifications. When a cluster instance is created this service will add MEX-specific applications to this service. Currently it creates prometheus-operator services whenever a cluster service is created.

• prometheus-operator - is a metrics collection framework and periodically scrapes the details from k8s. It is deployed as a helm chart

cluster-svc uses controller apis to create edgeproto.App and edgeproto.AppInst for the services using a pre-created flavor and developer

Usage

A typical usage of cluster-svc is as a docker container running on the MEX Platform. For an example see edge-cloud/e2e-tests/setups/mexdemo/mex-cluster-svc-deploy.yml

MCHU-MAC:edge-cloud matthewchu$ cluster-svc -h
Usage of cluster-svc:
  -ctrlAddrs string
    	address to connect to (default "127.0.0.1:55001")
  -d string
    	comma separated list of [etcd api notify dmedb dmereq locapi mexos metrics upgrade]
  -flavor string
    	App flavor for cluster-svc applications (default "x1.medium")
  -notifyAddrs string
    	Comma separated list of controller notify listener addresses (default "127.0.0.1:50001")
  -prometheus-ports string
    	ports to expose in form "tcp:123,udp:123"
  -r string
    	root directory for testing
  -scrapeInterval duration
    	Metrics collection interval (default 15s)
  -standalone
    	Standalone mode. AppInst data is pre-populated. Dme does not interact with controller. AppInsts can be created directly on Dme using controller AppInst API
  -tls string
    	server tls cert file.  Keyfile and CA file mex-ca.crt must be in same directory

TODO

1. scrapeInterval should be passed along to Prometheus app helm chart at creation time - currently defaults to 15secs
2. Need to create a separate flavor for infrastructure services not to conflict with user applications
3. Need to pass cloudlet and cluster information only when cluster is created to createAppInstCommon
   • EDGECLOUD-386
   • this will allow us to support multiple clusters/cloudlets