Closed d70-t closed 1 year ago
I just ran two builds in a row (first and second time).
I just re-run the build, and as #102 is now fixed, the book was built successfully, and used the cache for execution.
How does the new
save execution cache
fit in here now?
The issue here is, the default caching step will only upload the cache if the job was successful. The extra save execution cache
step also uploads the cache if the job failed (which I guess is what we need as failing jobs due to a single problematic notebook are frequent).
This PR should setup github actions such that the notebook execution cache will always be saved (even in case of errors). To my understanding, jupyterbook will re-execute failed notebooks on the next build, but could reuse successfully executed notebook cells in that case. This should therefore give use a big improvement in average build time.