A lightweight, multi-tenant, scalable and secure gateway that enables Jupyter Notebooks to share resources across distributed clusters such as Apache Spark, Kubernetes and others.
The CI tests have started failing the last few days[*][**] with no related changes (so probably a dependency issue). The problem is that I can't reproduce the failures, even with similar package versions, etc. I noticed that the docker log collection will not occur if the integration tests (which rely on docker) don't succeed. This change will perform that step regardless of the previous step's outcome.
Note, however, that should the unit tests fail (make test), the integration tests are not run and, as a result, the log collection step will fail since there's no docker instance from which to collect logs. We could go ahead and also run the integration tests unconditionally, but I felt that was getting us further away from a "fail fast" model, so choose to live with the log collection failures in this case.
[*]: This pull request also caps pyzmq < 24.0.0. This version was produced a few days ago and it appears to side affect the tests so we'll need to look into it. There was a 24.0.1 version immediately after the release, but that is causing the failure (or at least has some influence).
[**]: It turns out the pyzmq pin was a red-herring and the real culprit is the jupyter_server 1.19 release where the type of Future being used was changed. As a result, the pyzmq cap has been removed and a cap for jupyter_server<1.19 has been added for now.
The CI tests have started failing the last few days[*][**] with no related changes (so probably a dependency issue). The problem is that I can't reproduce the failures, even with similar package versions, etc. I noticed that the docker log collection will not occur if the integration tests (which rely on docker) don't succeed. This change will perform that step regardless of the previous step's outcome.
Note, however, that should the unit tests fail (
make test
), the integration tests are not run and, as a result, the log collection step will fail since there's no docker instance from which to collect logs. We could go ahead and also run the integration tests unconditionally, but I felt that was getting us further away from a "fail fast" model, so choose to live with the log collection failures in this case.[*]: This pull request also caps
pyzmq
< 24.0.0. This version was produced a few days ago and it appears to side affect the tests so we'll need to look into it. There was a 24.0.1 version immediately after the release, but that is causing the failure (or at least has some influence). [**]: It turns out thepyzmq
pin was a red-herring and the real culprit is the jupyter_server 1.19 release where the type ofFuture
being used was changed. As a result, thepyzmq
cap has been removed and a cap forjupyter_server<1.19
has been added for now.