Kubernetes - What happens if startupProbe runs beyond periodSeconds

Question

I have a Deployment which runs a simple apache server. I want to execute some commands after the service is up. I am not quite sure how much time the post action commands going to take. I have "timeoutSeconds" set as more than "periodSeconds".

Kube.nets Version: 1.25

apiVersion: apps/v1
kind: Deployment
metadata:
  name: readiness
spec:
  replicas: 1
  selector:
    matchLabels:
      app: readiness
  template:
    metadata:
      labels:
        app: readiness
    spec:
      containers:
        - image: sujeetkp/readiness:3.0
          name: readiness
          resources:
            limits:
              memory: "500M"
              cpu: "1"
          readinessProbe:
            httpGet:
              path: /health_monitor
              port: 80
            initialDelaySeconds: 20
            timeoutSeconds: 10
            failureThreshold: 20
            periodSeconds: 10
          livenessProbe:
            httpGet:
              path: /health_monitor
              port: 80
            initialDelaySeconds: 60
            timeoutSeconds: 10
            failureThreshold: 20
            periodSeconds: 10
          startupProbe:
            exec:
              command:
                - /bin/sh
                - -c
                - |-                 
                  OUTPUT=$(curl -s -o /dev/null -w %{http_code} http://localhost:80/health_monitor)
                  if [ $? -eq 0 ] && [ $OUTPUT -ge 200 ] && [ $OUTPUT -lt 400 ]
                  then 
                    echo "Success" >> /tmp/post_action_track
                    if [ ! -f /tmp/post_action_success ]
                    then
                      # Trigger Post Action
                      sleep 60
                      echo "Success" >> /tmp/post_action_success
                    fi
                  else 
                    exit 1
                  fi
            initialDelaySeconds: 20
            timeoutSeconds: 80
            failureThreshold: 20
            periodSeconds: 10

When I run this code, I see very strange results. As "periodSeconds" is 10 and my script has a sleep of 60 seconds, should not the start up probe trigger atleast 6 times, but it only triggers 2 times. I am checking the contents of files /tmp/post_action_success and /tmp/post_action_track to identify how many times the probe triggers. (Count the number of success inside the files)

Question: If the previous instance of startup probe is running, then is the startupProbe triggered on top of it or not? If yes, then why it triggered only twice in my case.

Another observation: When I set below options

initialDelaySeconds: 20
timeoutSeconds: 5
failureThreshold: 20
periodSeconds: 10

Then the content of file /tmp/post_action_success shows sleep/timeoutSeconds (60/5)=12 "success". Can someone please explain how this works.

Answer 1

I think the reason you see the probe being triggered twice is because of timeoutSeconds: 80 . See this question . Also the official doc is quiet handy in explaining the other fields.

Perhaps you can set initialDelaySeconds: 61 instead of using sleep in you script?