tl;dr - I set up a single instance of statping along with configuration to monitor uptime for a few sites I maintain.

I’ve written a bunch of times in the past about my k8s cluster (most posts are categorized under “k8s” or “kubernetes”), but during a recent rebuild of my cluster I forgot to put a rather important back up. It’s pretty disappointing that this happened at all – I went through the hard work of making all my deployments single-command (utilizing my still-possibly-a-terrible-idea makeinfra pattern), but I don’t have all the projects and more infrastructure related concerns in the same respository, which meant I had to go into each individual -infra repository for my other projects and run make deploy or something similar. For example, cert-manager (fantastic tool for automating certificate management on your cluster) is in the bigger “infrastructure” (not tied to any project), and projects like this blog have their own <project>-infra repositories.

Unfortunately the project I forgot to redeploy was actually the landing page to one of my businesses – gaisma.co.jp, and it actually wasn’t up when a client went to look at it. Of course, I had no idea the site was down, so when I pointed them to it they noted that it was down. Quite an unwelcome surprise!

So there are a few things I can do to stop this from happening again:

• git submodule the other projects into the main infra repo
• setup some site monitoring software in the overarching infrastructure repo

While the first solution is a good one (and probably something I should do period), the second one is what this post is going to be about. I’m going to set up statping which is a really nice looking self-hostable tool for simple uptime checks on one or more websites.

The real usefulness comes from automated downtime emails, which I’m going to hook up with my Gandi-provided free email and use to email myself when any sites go down.

Kubernetes

The deployment is pretty simple – I only need one instance of statping for the whole cluster (so no DaemonSet), It’s relatively “stateless” (so no StatefulSet), and I really only need one replica (so technically I could go with a Pod). I do want to be able to visit some status.<domain>.<tld> endpoint from the web and see everything at a glance, so I’ll need a Service and an Ingress.

Here’s what that looks like:

statping/pvc.yaml

---
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: statping-pvc
  namespace: monitoring
  labels:
    app: statping
    tier: monitoring
spec:
  storageClassName: openebs-jiva-non-ha
  accessModes:
  - ReadWriteOnce
  resources:
    requests:
      storage: 100Mi

statping/deployment.yaml

---
apiVersion: apps/v1
kind: Deployment
metadata:
  name: statping
  namespace: monitoring
spec:
  replicas: 1
  selector:
    matchLabels:
      app: statping
      tier: monitoring
  template:
    metadata:
      labels:
        app: statping
        tier: monitoring
    spec:
      containers:
        - name: statping
          image: hunterlong/statping:v0.80.51
          resources:
            limits:
              memory: 128Mi
              cpu: 100m
          ports:
          - containerPort: 8080
          volumeMounts:
            - mountPath: /app
              name: statping-config
      volumes:
        - name: statping-config
          persistentVolumeClaim:
            claimName: statping-pvc

statping/svc.yaml

---
apiVersion: v1
kind: Service
metadata:
  name: statping
  namespace: monitoring
spec:
  selector:
    app: statping
    tier: monitoring
  ports:
    - protocol: TCP
      port: 80
      targetPort: 8080

statping/ingress.yaml

---
apiVersion: extensions/v1beta1
kind: Ingress
metadata:
  name: statping
  namespace: monitoring
  annotations:
    ingress.kubernetes.io/ssl-redirect: "true"
    ingress.kubernetes.io/limit-rps: "20"
    ingress.kubernetes.io/proxy-body-size: "10m"
    kubernetes.io/tls-acme: "true"
    kubernetes.io/ingress.class: "traefik"
spec:
  tls:
  - hosts:
      - subdomain.domain.tld
    secretName: statping-tls
  rules:
  - host: subdomain.domain.tld
    http:
      paths:
      - path: /
        backend:
          serviceName: statping
          servicePort: 80

Another project that I was hoping to use with this is postmgr (which is almost at version 0.1.0!) – I created postmgr to make it easier to run well-configured Postfix servers but right now postmgr isn’t quite ready for dogfooding (it could stand to be a little more secure/featureful), so for now I’m deferring so some other cluster-external SMTP server (Gandi in this case).

After some make executions (which ran kubectl apply -fs), and some initial set up I was greeted with a nice uptime dashboard:

Check out more about statping at the Github Repo (thanks to hunterlong and all the contributors for making this awesome tool)! I also filed an issue asking about some documentation on config.yml, since I really wanted to be able to pre-populate the config.yml file with a configmap.

This post was a quick one but hopefully it shows how easy it is to deploy services with k8s and achieve some business value.