Airflow helm chart9/10/2023 ![]() ![]() But looks like Airflow is going to build upon the chart donated by Astronomer, and the stable chart will move into the bitnami one. I also found these discussions about moving the stable chart into Airflow itself. There were three main options: bitnami, stable, and the one by Astronomer. From the hub, I was able to launch a JupyterLab server which came up in its own pod!įinally, I teared everything down by just deleting the namespace.įor Airflow, I was confused about which public helm chart to use as I'd been using my own at my last job. NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE I re-ran the helm upgrade, and this time the hub pod came up nicely! When the hub was READY, I got the cluster IP for the proxy-public service: NAME STATUS VOLUME CAPACITY ACCESS MODES STORAGECLASS AGE ![]() microk8s-hostpath createdĬ8s.io/microk8s-hostpath createdĬ8s.io/microk8s-hostpath created I hadn't thought about storage classes on my smol laptop, but I found that you can enable a "default storage class" in microk8s.ĭeployment.apps/hostpath-provisioner created Normal FailedBinding 41s (x26 over 6m46s) persistentvolume-controller no persistent volumes available for this claim and no storage class is set $ microk8s kubectl describe pvc hub-db-dir -n jhub A kubectl describe on the PVC showed me that it wasn't able to come up because no storage class was set! Storage classes describe the different types of storage that your Kubernetes cluster has to offer, which are usually tied to your cloud provider's storage offerings (for example: EBS on AWS). The hub pod needs a PVC to store runtime metadata (like names of people who log in) in a sqlite db. Warning FailedScheduling 81s (x5 over 5m25s) default-scheduler 0/1 nodes are available: 1 pod has unbound immediate PersistentVolumeClaims. $ microk8s kubectl describe pod hub-766b95d6f9-vw9v4 -n jhub That showed me that it was still waiting for its PersistentVolumeClaim to come up. ![]() The deployments came up, but after the hub pod was stuck in PENDING state for a long time, I decided to do a kubectl describe on it. $ RELEASE=jhub NAMESPACE=jhub microk8s helm3 upgrade -cleanup-on-fail -install $RELEASE jupyterhub/jupyterhub -namespace $NAMESPACE -version=0.9.0 -values values.yml The docs mention a -create-namespace option but since that didn't work for me, I had to create the namespace manually using kubectl: The second step is to install the chart on your Kubernetes cluster using helm. Successfully got an update from the "jupyterhub" chart repository Hang tight while we grab the latest from your chart repositories. "jupyterhub" has been added to your repositoriesĪnd update all the repos to get the latest chart versions. The first step is to add the repo for the official JupyterHub helm chart: You can just enable it!Īfter enabling helm, I followed the steps in the awesome Zero to JupyterHub docs to set up JupyterHub. Also, you don't need to install helm separately. It comes pre-packaged with kubectl which you can invoke with microk8s kubectl. Microk8s v1.19.0 from Canonical* installed Microk8s is very easy to install on Ubuntu (No wonder! Both tools come from Canonical): Today I gave it a try and found those words to be true! In this post, I'll list down the steps (for future me) to set up working JupyterHub and Airflow instances on microk8s. Nemo had mentioned that microk8s is a lot better than minikube, and has perfected the developer experience. Till now, I haven't bothered installing anything on it just because of these long wait times. ![]() After a minikube start, it has been taking a long time to come up for me (maybe because of my smol laptop), and it also throws some errors before it finally comes up. Some days ago I set up minikube to put together a small demo (or at least some screenshots) for a talk. Day 24 - JupyterHub and Airflow on microk8s 10 September 2020 ![]()
0 Comments
Leave a Reply.AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |