Propagate start errors. #94
Conversation
jupyter_dash/jupyter_app.py
Outdated
| pass | ||
| req = requests.get(alive_url) | ||
| res = req.content.decode() | ||
| if req.status_code == 500: |
There was a problem hiding this comment.
Two questions here:
(1) Any chance we'd fail with a code other than 500? Maybe switch to if req.status_code != 200?
(2) Is there any way an error would make it into the queue later, ie during the requests.get call? Should we check the queue again after receiving this 500 response (or even check always after requests.get?
There was a problem hiding this comment.
(1) The error handler always return 500, that catches the errors in the before_first_request in dash and layout/deps errors. I don't think in dash we return other status code, but there could be with plugin.
(2) The queue get the error from server start, the requests error goes into Exception handler to return a 500, if the server failed to start the requests.get will raise connection error, next retry it get queue error if it wasn't there. Think if there a requests error it should check the queue after to ensure no race condition to get the proper error on the last retry.
Propagate start errors instead of the same
Address already in useerror messages when starting the threaded server.