by

Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Repo info
Activity
  • 15:20
    betatim commented #1123
  • 15:19
    betatim commented #1123
  • 15:01
    welcome[bot] commented #1123
  • 15:01
    fm75 labeled #1123
  • 15:01
    fm75 opened #1123
  • 14:34
    sgibson91 commented #1485
  • 10:15
    sgibson91 commented #1469
  • 10:09
    betatim commented #930
  • 09:40
    minrk commented #1469
  • 09:09
    betatim commented #930
  • 08:41
    fabiencelier commented #850
  • 08:08
    betatim opened #930
  • 07:59
    betatim commented #929
  • 07:53
    manics commented #929
  • 07:51
    betatim commented #929
  • 07:12
    betatim opened #929
  • 06:13
    betatim commented #925
  • 06:11
    betatim commented #925
  • Jul 08 23:27
    daroczig commented #773
  • Jul 08 16:09
    bitnik synchronize #927
Tim Head
@betatim
im not aware of any clusters on 1.18
Jeremy Tuloup
@jtpio
ok thanks @betatim :thumbsup:
a few experiments with 1.18 seem to be showing some notable differences: https://github.com/jupyterhub/binderhub/issues/810#issuecomment-651089134
Jeremy Tuloup
@jtpio

a few experiments with 1.18 seem to be showing some notable differences: https://github.com/jupyterhub/binderhub/issues/810#issuecomment-651089134

opened jupyterhub/binderhub#1116

Martin
@hackfin
Hope this is not an FAQ, i've got an /edit/ link in a NB that resolves on my local container (/edit/work/examples/vhdl/fifobuf.vhdl) but not on binder (where I get an URL like /user/hackfin-<magicnumber>/edit/...
Is there a known trick to resolve that?
Kenan Erdogan
@bitnik
@jtpio @betatim i recently upgraded GESIS cluster to 1.18.3 and so far i didnt notice any problems. i will look at the issue today
btw our cluster created on baremetal with kubeadm
Jeremy Tuloup
@jtpio
thanks @bitnik that's good to know :thumbsup:
Tim Head
@betatim
whoop!
Wayne's Bioinformatics Code Portal
@fomightez
Cool intersection highlighted in a twitter post: https://twitter.com/fperez_org/status/1278401194088517632
Uwe L. Korn
@xhochy
I'm making an internal presentation about how I deployed BinderHub and would love to show a small architecture diagram how Binder, JupyterHub, repo2docker, .. play together. Is there already one available from previous talks that I could use?
Sarah Gibson
@sgibson91
@xhochy The Turing Way have a live scribe sketch that I commissioned for these purposes, available under a CC-BY licence. Though I wouldn't call it an architecture diagram
BinderHub.jpg -> https://zenodo.org/record/3695300
Uwe L. Korn
@xhochy
Massive! That's so much better than a boring architecture diagram!
Sarah Gibson
@sgibson91
:blush:
Uwe L. Korn
@xhochy
(because it's accessible to system engineers and analytical people at the same time)
Sarah Gibson
@sgibson91
That's precisely why I asked for it! Working my way through the docs and figuring all this out was the best part of 6 months when I first joined the project. Anything to help reduce that barrier.
For completeness, there is also this: https://binderhub.readthedocs.io/en/latest/overview.html?highlight=architecture#a-diagram-of-the-binderhub-architecture
Chris Holdgraf
@choldgraf
✨✨✨✨
y'all are awesome!
Carol Willing
@willingc
Great conversation here folks :D
Philip Durbin
@pdurbin
Heh. Hi Carol. I'd sit at this computer a little longer except that we just got a dog and it's time to play. Hope you're well!
Fabien Celier
@fabiencelier
Hello guys,
I try to redirect to a custom URL on my binder but it looks like it fails because it adds the token at the wrong place: https://mybinder.org/v2/gh/fabiencelier/notebooks/master?urlpath=%2Fproxy%2F9999%2F%23%2Fdashboard%2Fab4
Would you have any idea how to solve that ? or an issue I can refer to ?
Kenan Erdogan
@bitnik
@fabiencelier token is ignored, because your urlpath contains '#'. i think this issue is related: jupyterhub/binderhub#850
Fabien Celier
@fabiencelier
Yes, that's my issue, thanks for the link :)
thewoz87
@thewoz87
Hello guys :sunglasses:
I've just installed binderhub on my openshift instance and everything is working properly except the final push . I receive the following error
"Error during build: 500 Server Error: Internal Server Error ("You cannot push a "root" repository. Please rename your repository to docker.io/<user>/<repo> (ex: docker.io/<token>/test-5f-2dogctb16-2deoap-2dd169-5f-2djupyter-5f-2dnb-2drabbit-5f-2dtest-35cd2c)")"
Can you help me?I tried modifying the config.yaml adding "image_prefix: thewoz87/bindertest" (it is a public repo on dockerhub) but the problem persist and honestly I don't understand what I have to modify....thx in advance for your support :)
Sarah Gibson
@sgibson91
@thewoz87 image prefix shouldn't be a repo, it's something like "binder-dev-" so you know where the image came from. Binder will actually create a new repo in DockerHub for every image it builds
Ah sorry
It should be image_prefix: <YOUR DOCKER USERNAME OR ORG>/<SOME TAG>
I'd double check your docker credentials are correct
thewoz87
@thewoz87
@sgibson91 Hello Sarah, thanks for your answer. Currently within my config.yaml I have " image_prefix: thewoz87/bindertest" . My credentials are correct in fact I can push manually a docker without any problem
Sarah Gibson
@sgibson91
Also, is may be better to post this as a question on our forum (https://discourse.jupyter.org) with your config (secrets redacted). The answers will be persistent and the config will help us debug :)
thewoz87
@thewoz87
ok I'll post my question here https://discourse.jupyter.org
thx
thewoz87
@thewoz87
done :ok_hand:
Sarah Gibson
@sgibson91
fab, thank you!
Fabien Celier
@fabiencelier
With mybinder I generally have to wait between 30 secondes and 1 minute to launch my server (when the image is found). Can I expect better performance if I deploy my own BinderHub ? How long do you guys usually wait ? What factors impact this launch time ?
Tim Head
@betatim
https://discourse.jupyter.org/t/how-to-reduce-mybinder-org-repository-startup-time/4956/ is a good thread on that
it depends on how large your docker image is as we (most likely) have to transfer it from the registry to a node
even if the image is on the node that it needs to be before you request a launch it can happen that we "lose" ~10s (startup takes 10s longer) because of a reason we haven't tracked down
so the best case launch time is about 15-17s
Fabien Celier
@fabiencelier
interesting link :)
Sarah Gibson
@sgibson91
I wonder if we should put that link on the loading page?
Fabien Celier
@fabiencelier
So if the image is already on the machine (which I expect to be on my own BinderHub) all the time will be spend in steps 1. (reserve a pod) and step 4 (launch the user’s session)
Tim Head
@betatim
yes
Fabien Celier
@fabiencelier
Is reserving the pod taking most of the time ? I can start a docker with jupyter and everything in a very short time on my own machine. Or is it this 10s delay you haven't track down ?
Tim Head
@betatim
it takes 1-2s for kubernetes to schedule the pod
1 reply
a small amount of time to start the "init pods to limit network access". a few seconds for the jupyter process to start, a few seconds for binderhub to notice, finally your browser needs to follow the redirect
so i think voerall there are a lot of small steps that take a very short amount of time plus a small delay but once you add them up you end up with 10s or so
then there is a ~10s delay sometimes on top of that which we haven't chased down