Pin nbconvert version for CI#9341
Conversation
|
Thank you for opening a new pull request. Before your PR can be merged it will first need to pass continuous integration tests and be reviewed. Sometimes the review process can be slow, so please be patient. While you're waiting, please feel free to review other open PRs. While only a subset of people are authorized to approve pull requests for merging, everyone is encouraged to review open pull requests. Doing reviews helps reduce the burden on the core team and helps make the project's code better for everyone. One or more of the the following people are requested to review this:
|
Since late December we've started seeing frequent job timeouts in the docs builds when it's running jupyter-execute cells as part of the docs builds. Looking at likely causes there were several package releases around that time (as well as a new ubuntu version which was ruled out as the root cause in Qiskit#9340) In attempt to debug and potentially prevent this (in the short term) failure mode this commit pins the version of ipykernel which is a likely potential cause of the failure. We can use this commit to test CI to see if the timeouts are reproducible and if we do not encounter the docs job timeout after multiple attempts we can merge this to hopefully prevent it in CI.
mtreinish
force-pushed
the
pin-ipykernel
branch
from
25ce3f6 to
6538ded
Compare
Pull Request Test Coverage Report for Build 3846730700
💛 - Coveralls |
It looks like the numpy pin fixes this, so lets revert the ipykernel version cap and see if running with just a pinned numpy is enough to fix this. This reverts commit 6538ded.
mtreinish
added
the
Changelog: None
|
Sigh, it looks like pinning numpy didn't fix this either. Since #9340 ruled out the move to ubuntu 22.04 this PR has confirmed that numpy and ipykernel releases didn't fix this. Since this failure mode was first reported on the Dec 22nd looking at releases of dependencies before then there are two likely candidates left, |
|
Oh, we already have ipywidgets pinned to < 8.0.3 so it probably won't be that (unless we're getting a weird cross interaction between old ipywidgets and something else). |
Pinning numpy didn't do the trick either. But looking at the full dependency tree (not just direct dependencies) the full jupyter stack released around the time the failure started. This commit pins nbconvert which is the most directly involved jupyter package for the jupyter-execute sphinx extension. If this doesn't work I will try unpibning ipywidgets and jupyter-core and suppressing the deprecation warnings about seaborn (if they still are an issue) to see if are partial jupyter stack pinning is the source of the hang.
|
So I forgot to run my script on the full dependency tree and was just running it on the direct dependencies. There were more packages released in the rough time frame that are involved in jupyter sphinx which are 2nd or 3rd level requirements. Primarily the full jupyter stack released and any of those packages are possible causes here. I've pinned nbconvert as a a new attempt to see if we can reproduce the failure. If this doesn't work I'll try removing our existing jupyter stack pins to see if there is an incompatibility between versions (since I expect given the number of packages jupyter only tests the latest versions all at once). |
|
Superseded by #9346 |
4 participants
Summary
Since late December we've started seeing frequent job timeouts in the docs builds when it's running jupyter-execute cells as part of the docs builds. Looking at likely causes there were several package releases around that time (as well as a new ubuntu version which was ruled out as the root cause in #9340) In attempt to debug and potentially prevent this (in the short term) failure mode this commit pins dependencies to try and reproduce the failure and see if we can avoid the
stuck docs build by holding back a dependency version.
Details and comments