Why my application URL works alternatively inside the Pod

An Attribute of ROSA Network Load Balancer

Created By Meta AI

I am having my application running on ROSA. A cronjob needs to run to perform the backup task. It needs to access the exposed application URL. However its noticed that the URL is only accessible alternatively. Sometimes, I can connect and get the result but sometime the connection timeout. While if the access is from external, outside from the pods, it is just working fine.

Duplicate the Problem

Lets duplicate the problem by running a pod, where we can run tcpdump to analysis the issue.

Create a service account and assign the privileged SCC.

Run the debug pod with network troubleshooting tool, nicolaka/netshoot.

apiVersion: v1
kind: Pod
metadata:
  labels:
    run: debug-pod
  name: debug-pod
spec:
  containers:
  - image: nicolaka/netshoot
    name: netshoot
    command:
    - bash
    - -c
    - "sleep 3600"
    securityContext:
      privileged: true
  dnsPolicy: ClusterFirst
  restartPolicy: Always
  serviceAccountName: debug-sa

Run the tcpdump tool and save the result into file. Concurrently exec into the pod and simulate the problem by running

curl -kLv…