How to change running pods limits in Kubernetes?

Question

I have a self made Kubernetes cluster consisting of VMs. My problem is, that the coredns pods are always go in CrashLoopBackOff state, and after a while they go back to Running as nothing happened.. One solution that I found and could not try yet, is changing the default memory limit from 170Mi to something higher. As I'm not an expert in this, I thought this is not a hard thing, but I don't know how to change a running pod's configuration. It may be impossible, but there must be a way to recreate them with new configuration. I tried with kubectl patch, and looked up rolling-update too, but I just can't figure it out. How can I change the limit?

Here is the relevant part of the pod's data:

apiVersion: v1
kind: Pod
metadata:
  annotations:
    cni.projectcalico.org/podIP: 176.16.0.12/32
  creationTimestamp: 2018-11-18T10:29:53Z
  generateName: coredns-78fcdf6894-
  labels:
    k8s-app: kube-dns
    pod-template-hash: "3497892450"
  name: coredns-78fcdf6894-gnlqw
  namespace: kube-system
  ownerReferences:
  - apiVersion: apps/v1
    blockOwnerDeletion: true
    controller: true
    kind: ReplicaSet
    name: coredns-78fcdf6894
    uid: e3349719-eb1c-11e8-9000-080027bbdf83
  resourceVersion: "73564"
  selfLink: /api/v1/namespaces/kube-system/pods/coredns-78fcdf6894-gnlqw
  uid: e34930db-eb1c-11e8-9000-080027bbdf83
spec:
  containers:
  - args:
    - -conf
    - /etc/coredns/Corefile
  image: k8s.gcr.io/coredns:1.1.3
  imagePullPolicy: IfNotPresent
  livenessProbe:
    failureThreshold: 5
    httpGet:
      path: /health
      port: 8080
      scheme: HTTP
    initialDelaySeconds: 60
    periodSeconds: 10
    successThreshold: 1
    timeoutSeconds: 5
  name: coredns
  ports:
  - containerPort: 53
    name: dns
    protocol: UDP
  - containerPort: 53
    name: dns-tcp
    protocol: TCP
  - containerPort: 9153
    name: metrics
    protocol: TCP
  resources:
    limits:
      memory: 170Mi
    requests:
      cpu: 100m
      memory: 70Mi

EDIT: It turned out, that in Ubuntu the Network Manager's dnsmasq drives the Corends pods crazy, so in /etc/NetworkManager/NetworkManager.conf I commented out the dnsmasq line, reboot and everything is okay.

Answer 1

You must edit coredns pod's template in coredns deployment definition:

kubectl edit deployment -n kube-system coredns

Once your default editor is opened with coredns deployment, in the templateSpec you will find part which is responsible for setting memory and cpu limits.