Ask quick dev questions about JupyterHub, the multi-user server for Jupyter notebooks. Use discourse.jupyter.org for user questions, support, and discussion.
Is it possible to use node from the other zone in jupyterhub?
I have a GKE cluster, located in europe-west3-a zone, all the node pools are also in the same zone. Now created a new node pool in europe-west3-b ( europe-west3-a doesn't support GPU).
However launching jhub on this node doesn't work (Autoscaler doesn't try to create a node in the new pool).
Does kubespawner specify specific location for nodes? Can it be overwritten?
I have a GKE cluster, located in europe-west3-a zone, all the node pools are also in the same zone. Now created a new node pool in europe-west3-b ( europe-west3-a doesn't support GPU).
However launching jhub on this node doesn't work (Autoscaler doesn't try to create a node in the new pool).
Does kubespawner specify specific location for nodes? Can it be overwritten?
What is the best way to run several version of python into the same jupyter instance ? I would like to test jupyter with pypy but I would like to avoid to install 2 jupyter with pip.
What is the best way to run several version of python into the same jupyter instance ? I would like to test jupyter with pypy but I would like to avoid to install 2 jupyter with pip.
One way may be to create a new conda environment with a different python version on the jupyter server. Then conda activate new_environment_name .
Hello. I've deployed JupyterHub on Kubernetes on a simple configuration. I've enabled the https but traefik cannot receive the ACME challenge. It is on GCP, with load balancer.
time="2020-07-02T22:04:22Z" level=info msg="Configuration loaded from file: /etc/traefik/traefik.toml"
time="2020-07-02T22:04:22Z" level=info msg="Traefik version 2.1.9 built on 2020-03-23T17:23:17Z"
time="2020-07-02T22:04:22Z" level=info msg="\nStats collection is disabled.\nHelp us improve Traefik by turning this feature on :)\nMore details on: https://docs.traefik.io/contributing/data-collection/\n"
time="2020-07-02T22:04:22Z" level=info msg="Starting provider aggregator.ProviderAggregator {}"
time="2020-07-02T22:04:22Z" level=info msg="Starting provider *file.Provider {\"watch\":true,\"filename\":\"/etc/traefik/dynamic.toml\"}"
time="2020-07-02T22:04:22Z" level=info msg="Starting provider *traefik.Provider {}"
time="2020-07-02T22:04:22Z" level=info msg="Starting provider *acme.Provider {\"email\":\"gillam@gmail.com\",\"caServer\":\"https://acme-v02.api.letsencrypt.org/directory\",\"storage\":\"/etc/acme/acme.json\",\"keyType\":\"RSA4096\",\"httpChallenge\":{\"entryPoint\":\"http\"},\"ResolverName\":\"le\",\"store\":{},\"ChallengeStore\":{}}"
time="2020-07-02T22:04:22Z" level=info msg="Testing certificate renew..." providerName=le.acme
time="2020-07-02T22:04:27Z" level=error msg="Unable to obtain ACME certificate for domains \"jupyter.healthlab.com\" : unable to generate a certificate for the domains [jupyter.healthlab.com]: acme: Error -> One or more domains had a problem:\n[jupyter.healthlab.com] acme: error: 400 :: urn:ietf:params:acme:error:connection :: Fetching http://jupyter.healthlab.com/.well-known/acme-challenge/RqpFfRG4Gu-NSVVbVG4WhE6Uj24tAmHXZtCPorl7324: Connection refused, url: \n" providerName=le.acme
Any ideas ?
Hello. I am running jhub on Kubernetes. For some reason, Helm gets deleted automatically from the server:
helm version only shows: version.BuildInfo{Version:"v3.2.1", GitCommit:"fe51cd1e31e6a764cba7dead9552a6d418ded79a", GitTreeState:"clean", GoVersion:"go1.13.10"} instead of both server and client versions. Jhub is working fine, but when I make a change in the yaml and update, I have to install Helm again with curl... before updating the Jhub. Any ideas appreciated..
6 replies
I have set up a JupyterHub system on a RHEL system. The server roughly works but it randomly stops responding for 1 to more than 30 minutes. Symptoms include no prompt for new terminal , kernel shows starting but not working for very long time. The server is powerful enough and I do not see any particular error from the log file (the process has been started with no error, just not working properly. Has anyone seen this? How can I debug it?
I've never seen anything like that @BoPeng . Have you tried the troubleshooting docs? https://jupyterhub.readthedocs.io/en/stable/troubleshooting.html
are you interested in running GL apps on a real GPU ??
We got this running in noVNC on nodes in Nautilus https://gitlab.com/nvidia/container-images/samples/-/blob/master/opengl/ubuntu16.04/turbovnc-virtualgl/Dockerfile
Thanks @kinow but none of the problems apply to us. It was very strange that sometimes the server would work ok, but sometimes very slowly. This looks more like a system problem (swap etc) but we could not identify anything and other applications (e.g. a Rstudio server on the same server) runs ok. I have turned on
--debug and will dig deeper.
any ideas what the issue could be? I'm using nginx -ingress, so I think that's where the problem lies
Like @kartoch - I'm having trouble with LetsEncrypt - with the challenge step - see : https://discourse.jupyter.org/t/trouble-getting-https-letsencrypt-working-with-0-9-0-beta-4/3583/3?u=matthew.brett
Hi all, I have deployed jupyterhub, my users want to use labs more than the old notebooks and the labs interface works fine.
How can I allow users to add extensions to their own lab? I can add server extensions but when a user tries they get a 500 error and in the logs it shows permission denied. I then opened up the permissions on the extensions folder, they were able to then install extensions but couldn't build the lab. I also noticed that if one user added an extension, they all got it...
I have found no documentation around any of this so would welcome some help please :)
Hi all! We're currently in the testing stage for deploying a jupyterhub for staff & students at our institute. So far things are working nicely.
One problem I do have yet, is how to get information about logged in users from the shell.
If I open a terminal as a user (from within jupyterhub), I will just see 'administrative users', when i issue anything like 'w' or 'who'. Funny enough this 'administrative user', say logged in through ssh, will NOT see the people using the hub at the moment. Well, they can be found through sideways (along ps aux | grep ...) -- but, is there something more straight forward?
The login is controlled through pam/ldap and works well!
One problem I do have yet, is how to get information about logged in users from the shell.
If I open a terminal as a user (from within jupyterhub), I will just see 'administrative users', when i issue anything like 'w' or 'who'. Funny enough this 'administrative user', say logged in through ssh, will NOT see the people using the hub at the moment. Well, they can be found through sideways (along ps aux | grep ...) -- but, is there something more straight forward?
The login is controlled through pam/ldap and works well!
[so the expected behaviour would be typing 'w' or 'who' on a non-jupyterhub shell and actually see who's logged in]
I have created kubernetes clusters in aws and i am trying to deploy using gitlab-ci.yaml. The helm and tiller are working normally when used from local machine or server. But when trying through runner, tiller gives the error:
Error: Get http://localhost:8080/api/v1/namespaces/kube-system/pods?labelSelector=app%3Dhelm%2Cname%3Dtiller: dial tcp 127.0.0.1:8080: connect: connection refused.
Has anyone faced this problem?
I am trying to write a spawner based in lxd/lxc containers. It creates a container for each user, is there a way to have a uid associated to each user?
Consider that the list of users is not known in advance: the authenticator might be anything, so putting a dictionary of uid in the config file is not an option
@consideRatio on jupyterhub/zero-to-jupyterhub-k8s#1708 rather than move everything to values.yaml and deprecate
.Values.hub.uid do you think it would be better to just allow for an optional .Values.hub.extraContainerSecurityContext value where people can put capabilities or whatever else so long as it doesn't touch what's defined in the upstream values.yaml, i.e. runAsUser?
1 reply
@minrk @BoPeng In our implementation, we used RedHat 7. port 443 is configured inside jupyterhuib_config.py. However, some apps will not show up at times, for example, if we requested multiple terminals, sometimes, all would show up, other times, none would. I wish there's a way to attach our configuration file here, or any other file. The log file reveals nothing useful - no red.
c.Application.log_level = 'DEBUG'
c.JupyterHub.cookie_secret_file = '/var/log/jupyterhub/jupyterhub_cookie_secret'
c.JupyterHub.db_url = 'sqlite:///var/log/jupyterhub/jupyterhub.sqlite'
c.JupyterHub.hub_ip = '10.15.1.25'
c.JupyterHub.ip = '10.15.1.25'
c.JupyterHub.pid_file = '/var/log/jupyterhub/jupyterhub.pid'
c.ConfigurableHTTPProxy.pid_file = '/var/log/jupyterhub/jupyterhub-proxy.pid'
c.JupyterHub.port = 443
import sys
c.JupyterHub.services = [
{
'name': 'cull-idle',
'admin': True,
'command': [sys.executable, '/etc/jupyterhub/cull_idle_servers.py', '--timeout=3000']
}
]
c.JupyterHub.ssl_cert = '/etc/pki/tls/certs/soshub.crt'
c.JupyterHub.ssl_key = '/etc/pki/tls/private/soshub.key'
c.Spawner.default_url = '/lab'
c.JupyterHub.cookie_secret_file = '/var/log/jupyterhub/jupyterhub_cookie_secret'
c.JupyterHub.db_url = 'sqlite:///var/log/jupyterhub/jupyterhub.sqlite'
c.JupyterHub.hub_ip = '10.15.1.25'
c.JupyterHub.ip = '10.15.1.25'
c.JupyterHub.pid_file = '/var/log/jupyterhub/jupyterhub.pid'
c.ConfigurableHTTPProxy.pid_file = '/var/log/jupyterhub/jupyterhub-proxy.pid'
c.JupyterHub.port = 443
import sys
c.JupyterHub.services = [
{
'name': 'cull-idle',
'admin': True,
'command': [sys.executable, '/etc/jupyterhub/cull_idle_servers.py', '--timeout=3000']
}
]
c.JupyterHub.ssl_cert = '/etc/pki/tls/certs/soshub.crt'
c.JupyterHub.ssl_key = '/etc/pki/tls/private/soshub.key'
c.Spawner.default_url = '/lab'
just uploaded useful lines, omitting comments from jupyterhub_config.py
thanks! Dakai
Is there a way to retrigger this travis build https://travis-ci.org/github/jupyterhub/zero-to-jupyterhub-k8s/jobs/705490542 ? It seems to have failed on an unrelated issue getting a cert.
2 replies
Good morning everyone, I have the following error, could someone help me please :)
1 reply
Hello. First time poster here. I'm trying to install JupyterHub on my local Docker Desktop Kubernetes installation, but I keep running into
pod has unbound immediate PersistentVolumeClaims. When I describe the relevant PVC, it says waiting for a volume to be created, either by external provisioner "docker.io/hostpath" or manually created by system administrator. Anyone run into this particular issue?
I gather that if I were deploying to a cloud provider, it would provision properly, but I want to setup a local development workflow, so providing compatibility with Docker Desktop Kubernetes would be nice.
Hello, do you know if this is possible to mix binder, jupyterhub for kubernetes and jupyter entreprise gateway. The idea is the following: every time someone starts a jupyter notebook, a docker image is built using repo2docker and then the jupyter notebook based on this image is served to the user. I found ways to connect jupyterhub for kubernetes and jupyter entreprise gateway but nothing about binder used like this way.
postStart is post start in parallel to the other entrypoint process i think, so perhaps doing that will be too late.
i have a wrapping helm chart that writes some files to a configmap, which i then mount with extraVolumes and extraVolumesMount, like the mybinder.org-deploy github-repo can demondtrate how to do
Hey gang, is it possible to use implicit grant auth flow (providing id/access tokens directly to jhub) instead of authorization code flow? Background: we're using AWS Cognito, which only supports auth code flow via Hosted UI, which is a clusterf---. Cognito does support implicit grant flow (but not auth code flow) without the Hosted UI, so my question is whether vanilla jhub is flexible enough to accommodate the rigidity of Cognito in this case.
@consideRatio does jupyterhub/zero-to-jupyterhub-k8s#1708 require a CHANGELOG entry or do you roll those up at release time using git history?
1 reply
