Closed post2web closed 8 years ago
@jhpedemonte is the current release of the dash server busted because of the Jupyter lab API break?
Ah, yes, hadn't considered that. The 0.7.0 release will probably pull down the latest Jupyterlab, which will break things, as @post2web has found. master is working -- we pinned Jupyterlab 0.1.1.
@post2web, try the master branch for the dashboard server.
Maybe not: 0.7.0 was using the separate jupyter-js-ui, etc. libs prior to the merge into the single jupyterlab package.
https://github.com/jupyter-incubator/dashboards_server/blob/0.7.x/package.json
@post2web Do you see any console errors (or interesting output) from the kernel gateway or the dashboard server when rendering a dashboard? What about errors in the browser console?
@post2web Also, check out the jupyter-incubator/dashboards_setup project. It contains several recipes to quickly start the kernel gateway, notebook server and dashboards server. In particular, check the docker_deploy folder.
@jhpedemonte There is no console errors. I tried the master and did not work also.
I tried starting the kernel gateway as shown here:
jupyter-kernelgateway --KernelGatewayApp.port=8889 --KernelGatewayApp.ip=127.0.0.1
But no luck. Notice when I do:
jupyter-dashboards-server --KERNEL_GATEWAY_URL=http://foo.com
It acts the same and no console errors.
To me it looks like the dashboard server can not communicate with a kernel and its not complaining about it.
OK, I figured it out. Since you are working from source, in addition to npm install
, you also have to run gulp build
. This will build the resources required by the browser.
However, there is a bug in a dependency of a dependency -- you'll have to run npm install jquery-ui@1.10.5
before doing gulp build
.
I'll create a new release (0.8.0), where these issues are fixed. I might also look to change the behavior of make build
to do both npm install
and gulp build
, to handle a scenario such as this.
NOTE: Running from source is covered in the readme. The relevant step is make dev
, which builds the client files before running the dashboard server.
Alternatively, you can install the actual built program using npm. That doesn't require that you run any make/build commands.
@jhpedemonte So the 0.7.0 is ok but master is busted?
No, 0.7.0 is broken because of dependency change, but master is working.
Ok. Should probably cut a 0.8 since it's going to be needed to bring up the VMs for the PyData Carolinas dashboard talk soon. Let me know if you want to do the release or want me to.
/cc @jtyberg @lbustelo
EDIT: retracted, see next comment
FYI, the stable 0.7.0 release works fine for me following the recipe from dashboards_setup/docker_deploy.
@post2web said:
To install the dashboard server I download and extract the 0.7.0 release, then run:
npm install -g jupyter-dashboards-server
should be all you need to do to get the latest stable release.
Closing. I released 0.8.0, which will fix building from source.
I only see the markdown cells on the notebook when the dashboad_server renders it. I have started a kernel gateway like so:
jupyter kernelgateway
To install the dashboard server I download and extract the 0.7.0 release, then run:To start:
./bin/jupyter-dashboards-server --KERNEL_GATEWAY_URL=http://127.0.0.1:8888
Then when I go to http://localhost:3000/dashboards/test none of the dashboards render the way they are supposed to.What am I doing wrong? Do I have to start the kernel gateway in a spacial way?