Skip to content

kubernetes / k8s

this is going to be a big one. see also kubectl

default hostname

this is for my future self, if you want to access services between deployments/pods:


get logs from all pods part of deployment

…use labels:

kubectl logs -f -l app=name

rancher, canal, flannel listen on network interface

I use wireguard to tunnel traffic between kubernetes nodes. When setting up a new rancher cluster, you can put in all your information in a GUI, set canal as the default network driver. afterwards edit the YAML part and go look for network. in there you can change the networking interface for the driver

    iface: "wg0"
    flannel_backend_type: "vxlan"
  plugin: "canal"

the iface part is important

re-deploy kubernetes pods / deployment

say, you have a deployment running and set it to Always pull an image, but to trigger this you need to redeploy it. let's also say, you're lazy and don't have a fully featured ci/cd pipeline for your raspberry pi kubernetes cluster at home. so let's just do this to re-deploy your stuff and force a re-download of the image

kubectl rollout restart deploy <deployment-name>

reisze aws-ebs

kubectl get storageclass
kubectl edit storageclass
  fsType: ext4
  type: gp2
allowVolumeExpansion: true

make note of the added allowVolumeExpansion.

  1. edit the volume claim
kubectl edit pvc gitlab-minio
  - ReadWriteOnce
      storage: 100Gi
  storageClassName: gp2
  volumeName: pvc-9ef67592-1d5e-11e9-857d-0a8271abe0a2

change the storage part.

  1. see status
kubectl describe pvc gitlab-redis
  Type       Status  LastProbeTime                     LastTransitionTime                Reason  Message
  ----       ------  -----------------                 ------------------                ------  -------
  Resizing   True    Mon, 01 Jan 0001 00:00:00 +0000   Wed, 23 Jan 2019 08:58:57 +0100           

it'll change to something like that hopefully:

  Type                      Status  LastProbeTime                     LastTransitionTime                Reason  Message
  ----                      ------  -----------------                 ------------------                ------  -------
  FileSystemResizePending   True    Mon, 01 Jan 0001 00:00:00 +0000   Wed, 23 Jan 2019 09:07:29 +0100           Waiting for user to (re-)start a pod to finish file system resize of volume on node.
Events:                     <none>
Mounted By:                 gitlab-redis-7b9d4587f8-v8jgw

afterwards you can delete the pod to re-create it (if you've got a replicaset hopefully to do that)

kubectl delete pod gitlab-redis-7b9d4587f8-v8jgw

run one-off cronjob immediately

kubectl create job --from=cronjob/your-configured-cron your-cron-manual

stop kubernetes deployment

set the replicas to 0, it'll delete the pods and the deployment will stay intact.

kubectl scale --replicas=0 deployment <deployment-name>

suspend all cronjobs

to disable / suspend all cronjobs, do this:

kubectl get cronjobs | grep False | cut -d' ' -f 1 | xargs kubectl patch cronjobs -p '{"spec" : {"suspend" : true }}'

re-enable like this:

kubectl get cronjobs | grep True | cut -d' ' -f 1 | xargs kubectl patch cronjobs -p '{"spec" : {"suspend" : false }}'


force-update images

if you're using images with the :latest tag or another tag which changes remotely but not by tag name/version, you can configure the imagePullPolicy to always pull the image when the pod is being deleted, or the deployment restarted.

Here's an example deployment:

kind: Deployment
        - image: example:latest
          imagePullPolicy: Always

run interactive container

useful for stupid annoying things when you just want a shell:

kubectl run alpine --image=alpine --rm -i --tty -- sh