Readthedocs.org: Build error with build id #8923859

Created on 16 Apr 2019  路  10Comments  路  Source: readthedocs/readthedocs.org

Details

The GeoPandas (https://github.com/geopandas/geopandas) doc building is already failing for some time. I assume this is due to the conda environment creation that is taking too resources.

Some time ago, I put some effort into improving this (eg I pinned all versions in the environment.yaml file, as this speeds up the environment creation), and then sometimes the build succeeded (like 1 in 10 times), but now it is again failing all the time.
I know https://docs.readthedocs.io/en/stable/guides/build-using-too-many-resources.html is indicating that it is a known problem with conda, and advising to use pip.

https://docs.readthedocs.io/en/stable/builds.html says "We can increase build limits on a per-project basis, if you provide a good reason your documentation needs more resources.". Therefore, I was wondering in which way this can be asked (here via github?)

Specifically for the docs of GeoPandas:

  • we use sphinx-gallery for the example gallery, which means we need to install the dependencies needed for running those examples, and not only just geopandas itself
  • for those dependencies (the "geospatial stack", i.e. gdal, proj, libspatialindex, etc and their python wrappers), it is difficult to use pip as those packages don't always provide wheels that include the necessary C library, and thus those need to be installed separately using a system package manager (I don't see any option in the config file to do a apt-get install ?).
    conda handles this natively as it just installs those as additional dependencies.
Support

Most helpful comment

Thanks a lot! Very much appreciated, and thanks for the awesome readthedocs platform

All 10 comments

This is somewhat blocking a new release (since I want to be sure to have docs for the released version).

Is there a possibility to manually upload html files for a certain tag? To at least have the docs for that release, making the actual issue less urgent?

We don't support uploading html files for now, see #1083. Hope we can increase the memory on monday.

@stsewd that would be great, thanks for considering it!

Sorry for pinging (I know this is not necessarily a priority for you), but any update?

I moved the Geopandas project to another queue, but it didn't help. So I ended up increasing the memory limits to 1200 mb. I just triggered a new build and it was successful https://readthedocs.org/projects/geopandas/builds/8960434/.

Let me know if you still have more problems.

Thanks a lot! Very much appreciated, and thanks for the awesome readthedocs platform

This started reappearing again a few days ago (conda giving memory problems):

$ conda env create --quiet --name stable --file doc/environment.yml
Collecting package metadata (repodata.json): ...working... done
Solving environment: ...working... done
Preparing transaction: ...working... done
Verifying transaction: ...working... Killed

Command killed due to excessive memory consumption 
Command time: 197s Return: 137 

I didn't note any change on our side (no additional dependencies, it's also failing on the stable docs that worked a few days ago), and also didn't see a conda change (eg new conda version) compared to the working builds of a few days ago. Is it possible something changed on readthedocs' end?

@jorisvandenbossche check if the python version wasn't updated, I saw some problems related to that when python 3.8 was released

Python is pinned to 3.7 in the environment.yml file, so that shouldn't have changed I think.

@jorisvandenbossche I've increased the memory limit to 1.5G https://readthedocs.org/projects/geopandas/builds/10018413/ Not sure why is using more resources now, maybe an increase of packages in the conda-forge channel?

Was this page helpful?
0 / 5 - 0 ratings

Related issues

dxgldotorg picture dxgldotorg  路  3Comments

davidfischer picture davidfischer  路  4Comments

adamjstewart picture adamjstewart  路  4Comments

SylvainCorlay picture SylvainCorlay  路  3Comments

PowerKiKi picture PowerKiKi  路  4Comments