Skip to content

Latest commit

 

History

History
64 lines (46 loc) · 1.42 KB

monitoring.md

File metadata and controls

64 lines (46 loc) · 1.42 KB
Raw

Kubernetes Monitoring

The following instructions are verified to work with kubernetes v1.11.2

Log in to master

cd ~
git clone https://github.com/5g-media/kubernetes-prometheus.git
cd kubernetes-prometheus
git checkout 7a0dd19d03c8909420ee1031ea5c5509cde134df
kubectl create namespace monitoring
kubectl create -f clusterRole.yaml
kubectl create -f config-map.yaml
kubectl create -f prometheus-deployment.yaml 

kubectl create -f prometheus-service.yaml --namespace=monitoring

Install Grafana

Login to master

docker run -d -p 3000:3000 --name grafana grafana/grafana:4.6.3

Configure Grafana

Login to UI: http://master_ip_address:3000(admin:admin)

Create data source

  • Type: Prometheus
  • HTTP Settings -> URL: http://master_ip_address:30000
  • Access: Proxy
  • Save and Test

Import dashboard

  • Dashboard -> Import
  • From json -> Paste contents of 5G MEDIA Apache OpenWhisk VIM-GPU-1536160005130.json

Install Kube State Metrics (to expose kube_pod_labels)

Log into master

cd ~
git clone https://github.com/kubernetes/kube-state-metrics.git
cd kube-state-metrics
git checkout 1b79b318b25f80aa1a771e5da7611c69a4888304
kubectl apply -f kubernetes

Add GPU Host metrics

TBD

Note

If, nevertheless, you are using a recent kubernetes cluster, pay attention to removed-cadvisor-metric-labels