Service domain name resolution not reliable

Jonas

12/28/23, 4:20 PM

This is Kubernetes service

apiVersion: v1
kind: Service
metadata:
  name: mysvc

spec:
  clusterIP: None
  
  selector:
    app: myapp
    type: main

  ports:
    - name: http
      appProtocol: http
      targetPort: http
      port: 8000

that serves 9 pods. Most of the time mysvc.default.svc.cluster.local resolves into 9 IP addresses. Randomly it resolves into 8 IP addresses (one IP address disappears) while pods stay the same and healthy. Pods are located on separate nodes that are located on multiple continents. There are 20 CoreDNS instances that are equally distributed on each node.

Is it possible to solve this issue?

Self-hosted Kubernetes 1.25.3

1 + 0

kubernetes

Score:1

Server

Jonas

12/29/23, 10:40 AM

Turns out that the issue was not related to CoreDNS. Pods became not ready for a short period of time. It happened because one of the pod containers was a bash script that was routinely exiting and starting again. Restart happened really fast but that was enough time for the restarting pod to get removed from CoreDNS domain resolution response.

+ 0

Elon Musk

I sit in a Tesla and translated this thread with Ai:

EN: Service domain name resolution not reliable

Post an answer

Most people don’t grasp that asking a lot of questions unlocks learning and improves interpersonal bonding. In Alison’s studies, for example, though people could accurately recall how many questions had been asked in their conversations, they didn’t intuit the link between questions and liking. Across four studies, in which participants were engaged in conversations themselves or read transcripts of others’ conversations, people tended not to realize that question asking would influence—or had influenced—the level of amity between the conversationalists.