Skip to content

Latest commit

 

History

History
111 lines (79 loc) · 3.14 KB

README.md

File metadata and controls

111 lines (79 loc) · 3.14 KB
Raw

Monitor Lab

We want to know the state of our cluster and make sure things stay up! In this lab we will monitor the status of our cluster.

Performance Monitoring

We will start by following the Heapster Installation Guide with some changes due to Metacloud.

get Heapster:

cd 
git clone https://github.com/kubernetes/heapster.git
cd heapster

Install several components

kubectl create -f deploy/kube-config/influxdb/

This will create 3 pods, heapster, graphana, and influxdb.

Ensure they are up:

kubectl get pods -n kube-system -o wide
NAME                                    READY     STATUS    RESTARTS   AGE       IP           NODE
heapster-2193675300-nktsw               1/1       Running   0          1m        10.210.1.3   vkw02
kube-dns-v20-1485703853-ah43b           3/3       Running   16         13h       10.210.0.2   vkw01
kube-dns-v20-1485703853-zmg5k           3/3       Running   0          13h       10.210.2.2   vkw03
kubernetes-dashboard-3203700628-x8mgw   1/1       Running   12         13h       10.210.1.2   vkw02
monitoring-grafana-927606581-wvgos      1/1       Running   0          1m        10.210.2.3   vkw03
monitoring-influxdb-3276295126-kalzd    1/1       Running   0          1m        10.210.0.3   vkw01

Now make sure our load balancer is pointing to this service.

kubectl edit -n kube-system svc monitoring-grafana

This will open a VI session. Modify nodePort: XXXX to be nodePort: 30861

This has been previously set up on the nginx cluster to reverse proxy to this port. Now you should be able to access your load balancer on port 3000 to see the grafana data.

Log Monitoring

1. Install Fluentd DaemonSet

fluentd is used for collecting container logs.

Run:

kubectl apply -f https://raw.githubusercontent.com/CiscoCloud/k8sclass/master/05-Monitor/fluentd-elasticsearch-v1-daemonset.yaml

This creates a Daemon Set which is essentially policy that states one pod of will run on every worker node in the cluster.

2. Install Elastisearch

We follow the guide shown in the Kuberenetes production cluster examples with some modifications to make it work on our cluster.

Run the following commands:

kubectl -f

Sources

fluentd daemon set

Troubleshooting

Create busybox host to test records. Make a yaml file called bb.yaml:

apiVersion: v1
kind: Pod
metadata:
  name: busybox
  namespace: default
spec:
  containers:
  - image: busybox
    command:
      - sleep
      - "3600"
    imagePullPolicy: IfNotPresent
    name: busybox
  restartPolicy: Always

Now run this:

kubectl apply -f bb.yaml

Now to log into it:

kubectl exec -it busybox -- /bin/sh

That will put you on the shell to run commands.

To see all the name spaces we have:

kubectl get svc,pods,deployments,daemonset --all-namespaces