For one of my production JupyterHub deployments on Kubernetes, I have setup an automated system to perform nightly backup of the user data, see the full tutorial on how to set it up.

The system writes the backups to Object store using Restic.

In this tutorial I’ll provide the configuration to have a CronJob on Kubernetes checking how old are the backups and be alerted if anything is not working using Healthchecks.io.

Healthchecks.io is a free service that gives you a URL endpoint you should regularly send a GET request to, for example from a bash script, if they don’t receive a ping after a configurable amount of hours, they send an email.

Setup the system

First register for a free account at https://healthchecks.io, configure an endpoint and get the related URL.

Checkout my usual zonca/jupyterhub-deploy-kubernetes-jetstream repository from Github, enter the backup_volumes folder.

Make sure you have the Restic password and the AWS-style credentials saved in text files in the same folder, make sure there is no newline at the end of the files (use vim -b with :set noeol, yes, this is a reminder for myself).

bash create_aws_secret.sh

Then enter the backup_volumes/monitor subfolder:

  • Edit cronjob.yaml and enter your Healthchecks.io URL
  • Run the backup_is_current.sh script locally to make sure it works properly
  • Run the create_configmap_scripts.sh bash script to load the previous script into Kubernetes as a configmap
  • Modify the schedule in cronjob.yaml to “* * * * *”, so every minute it has a chance to run correctly
  • Finally run kubectl apply -f cronjob.yaml to create the CronJob
  • Debug until it works!
  • If you login to Healthchecks.io, you should see the GET requests coming in every minute.
  • Relax, you’ll be notified if backups stop working for any reason.