Hacktoberfest 2026: le issue che i maintainer hanno segnato per ottobre, aperte e adatte ai principianti. Sfoglia le issue Hacktoberfest

Suggested doc improvements based on first-time user experience

Aperta
  • Qualcosa non va?
#129 1 commento 1 reazione 0 assegnatari Vedi su GitHub

Nessuno ha ancora preso questa issue.

Valutazione

Difficoltà
3/5
Tempo stimato
1-2 giorni
Idoneità per principianti
38/100
Tipo di issue
Documentazione
Chiarezza
Abbastanza chiara
Stato di attività
Ferma
Stack tecnologico
docker, helm, kubernetes, python
Ambito
cloud, devops, documentation

Direzione di ricerca

Inizia dal README del chart e dalla relativa guida rapida, quindi esamina il Dockerfile, environment.yml e le impostazioni dell’immagine documentati e menzionati nell’issue. Aggiorna la documentazione per spiegare le limitazioni del chart, la configurazione corretta di GatewayCluster, il workflow per le immagini personalizzate e il pinning delle versioni. Il lavoro è completato quando un utente GKE alle prime armi può seguire la guida senza dover fare affidamento sulle discussioni esterne collegate.

Scritto dal modello di indicizzazione a partire dal testo della issue.

Descrizione

Thank you for this fantastic daskhub chart. Myself and @saschahofmann has just set this up on Google Kubernetes Engine. I thought it would be helpful to share our experience, gotchas and potentially suggest improvements to the docs.

Making the decision between charts

We started with the simpler "dask" helm chart.

However it didn't really suit our need for a few reasons:

  • The jupyter notebook has no persistent disk storage
  • We wanted the ability to control the number of workers, without making updates via helm

It would have been helpful to have these limitations listed on the README, especially the persistence one which I'd imagine most people would expect (I guess will #78 fix it though).

Setting up daskhub

Setting up the initial "daskhub" worked fine, out of the box. The one part that is a bit flaky for us is the "launch cluster" button on the left sidebar. Sometimes it launches a local cluster, rather than via dask-gateway - we haven't worked out why yet.

Small thing - the quickstart does not quite work for us:

>>> from dask_gateway import GatewayCluster
>>> cluster = gateway.new_cluster()
>>> client = cluster.get_client()

Instead we ran:

from dask_gateway import GatewayCluster
# This line is missing
gateway = GatewayCluster()
cluster = gateway.new_cluster()
client = cluster.get_client()

Customising the image 🏗️

For our use case, we needed to build a custom image.

The default tag for daskhub is pangeo/base-notebook:{XXX}. Unfortunately, googling that led us to the pangeo stacks Github and webpage and spent time trying to get the "ONBUILD" images to work.

Thankfully we eventually stumbled on this comment, and led us to the correct repo. The "ONBUILD" trick is neat, just took us time to get our head around.

Anyone else stuck doing the same thing, here is what worked for us:

  • Create a local copy of the pangeo/base-notebook directory here. You minimally need the Dockerfile, apt.txt, environment.yml, postBuild and start.
  • Modify the apt.txt with system packages, and environment.yml with your extra conda packages
  • Docker build and push
  • Set this to the jupyterhub.singleuser.image.{name, tag, pullPolicy}

For the dask custom image, we just used daskgateway/dask-gateway we just did:

# Dockerfile
FROM daskgateway/dask-gateway:0.9.0
COPY ./environment.yml ./environment.yml
RUN conda env update -f environment.yml

# environment.yml
name: base
dependencies:
  - # extra packages...

And set this to dask-gateway.gateway.backend.image.{name, tag, pullPolicy}

Making sure to tag specific versions 🤦

A few days ago, we discovered inexplicably that we could no longer build new custom images that worked on the cluster. Turns out we hadn't pinned our images, and dask-gateway 0.9.0 was released.

Obviously completely our fault, but wanted to note that it's essential to have consistent versions for:

  • The helm chart
  • The pangeo/base-notebook in your jupyterhub.singleuser Dockerfile
  • The daskgateway/dask-gateway in your Dockerfile

Otherwise things will break in hard to debug ways. A particularly nasty example was using dask dataframe, where we had pandas 1.1.* on the client and 1.0.* on the workers.

We'd be happy to submit a PR to amend the docs or maybe a separate guide if that's useful, please let us know.

Once again, thanks for building this!

Lingua principale
YAML
Stelle
100
Fork
92
Nessuna PR unita negli ultimi 30g

Preparare l'ambiente

Come iniziare

  1. Leggi tutta la issue e poi la guida ai contributi del progetto.
  2. Commenta sulla issue per dire che te ne occupi tu — evita che due persone facciano lo stesso lavoro.
  3. Fai un fork del repository e lavora su un branch.
  4. Apri una pull request che faccia riferimento al numero della issue.

Altre issue di dask/helm-chart

Tutte le issue di dask/helm-chart

Issue simili

Altre issue su Cloud

Ricevi le nuove issue nella tua casella

Un breve riepilogo di issue GitHub adatte ai principianti.