Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Repo info
Activity
    Steven Silvester
    @blink1073
    does the demo work on your machine?
    Brian E. Granger
    @ellisonbg
    Yes and no
    I can do a small calculation (toy), but on the cluster we can actually do something very non-trivial - the visual+demo impact is much greater...
    Steven Silvester
    @blink1073
    I’ve never used either dask or a cluster
    Brian E. Granger
    @ellisonbg
    From our perspective, it should just be a regular jupyterlab installation
    The cluster stuff is all separate
    And matt will have dask already installed and ready to go
    The only part is getting jupyterlab installed with the dask_labextension
    The dask/bokeh part Matt knows how to setup
    Steven Silvester
    @blink1073
    which I have right now with the HEAD of both
    the extension loads on the page, but there’s no dask process to back it up
    Brian E. Granger
    @ellisonbg
    OK, then it should be working!
    Let's have @mrocklin give it a shot!
    Steven Silvester
    @blink1073
    it looks like we still have an install issue on his machine
    he can’t launch jlab
    Brian E. Granger
    @ellisonbg
    Ahh, what is the issue do you think?
    Steven Silvester
    @blink1073
    I don’t know yet without more debug back and forth
    I don’t know if he’s trying to run jlab on the cluster, and what problems that might introduce
    Brian E. Granger
    @ellisonbg
    No, he would just run jlab on a single public facing node
    The cluster stuff is all behind the scenes - only dask would know about that
    Other than the installation/build the only thing is to get the notebook to listen on a public IP/port
    But Matt already knows how to do that
    Steven Silvester
    @blink1073
    okay, then I’ll hold off until he gets back and try and help with the install issue
    Brian E. Granger
    @ellisonbg
    ok great
    @mrocklin my talk is at 1:15 in 1 E 07/1 E 08
    Can you come early - like 12:45 and we can test it?
    Steven Silvester
    @blink1073
    @ellisonbg, can you point me in the right direction to get it running locally, or it is not trivial?
    Brian E. Granger
    @ellisonbg
    I don't think that is worth it
    Steven Silvester
    @blink1073
    gotcha
    Brian E. Granger
    @ellisonbg
    This notebook works
    Run it and watch:
    Steven Silvester
    @blink1073
    no luck: ImportError: cannot import name ‘Client’, after having installed dask.distributed
    Brian E. Granger
    @ellisonbg
    Hmm, I literally just ran that notebook minutes ago...
    Steven Silvester
    @blink1073
    I found a workaround, proceeding
    Matthew Rocklin
    @mrocklin
    @ellisonbg where are you? Shall we meet briefly?
    Matthew Rocklin
    @mrocklin
    @ellisonbg I'm happy to meet up any time. Prefer sooner to later to make sure everything is smooth. Give me a call or text at 510-926-2779. I'm not checking in with Gitter as often as I'd like.
    Things are a bit crazy here
    To be clear, I'm not confident that I can get jlab running and accessible on the cluster.
    Brian E. Granger
    @ellisonbg
    OK, no problem
    Will ping you
    Will arrive there in 30m
    Matthew Rocklin
    @mrocklin
    OK, will meet you at the room
    Steven Silvester
    @blink1073
    I’ll be afk for lunch for the next hour, @mrocklin you can ping me on Flowdock if you have a question
    Matthew Rocklin
    @mrocklin
    Everything is done. Thanks for your high pressure efforts everyone.
    Luke Canavan
    @canavandl
    :clap: :clap:
    Pierre Glaser
    @pierreglaser
    Hi all!
    I have been using dask-jobqueue in jupyter lab to launch experiments on slurm clusters. So far, the experience is really pleasant.
    Pierre Glaser
    @pierreglaser
    One thing I've come across though: the dask_labextension is not happy if I run jupyter lab in another environment than the one I use to run my experiments. For the sake of isolation, I actually run jupyter (which includes all its tooling suite, including dask_labextension) in a separate environment, and manually add specific ipython kernels to it. However, when trying to launch a LocalCluster and connect to it using the dask_labextension UI (using jupyter lab from a different environment than the kernel I'm running code on), I receive an error message telling me that I should not use different dask versions at the same time.
    Am I doing something wrong? Is this pattern supported?