undefined symbol: mkl_serv_getenv in numexpr

[joshowen]

We are not using any MKL optimized libraries, but are getting this, where we didn't previously

python: symbol lookup error: /app/.heroku/miniconda/envs/heroku-env/lib/python2.7/site-packages/numexpr/../../../libmkl_vml_avx.so: undefined symbol: mkl_serv_getenv 

[ilanschnell]

This blog post: https://www.continuum.io/blog/developer-blog/anaconda-25-release-now-mkl-optimizations
Describes how to not use MKL linked versions of packages (on Unix). Nevertheless, the symbol error indicates that a MKL library is missing in the linkage. We have to look into that.

[ozancaglayan]

So somehow this morning conda brought to me the mkl package and I've hit the same problem. Now I did an update --all and it downgrades some numexpr, numpy, scipy and scikit-learn and installs mkl-rt package. And it's working correctly now.

[kmdouglass]

I have also received this error. One of my environments needed PyTables so I performed a conda install pytables, which also upgraded a package related to MKL (I can't remember what package specifically; if you could point me towards a relevant log file I can find it.)

After this, I would receive the following error message from a Jupyter Notebook session when performing matrix-heavy operations:

/home/douglass/anaconda3/envs/homogenizer/bin/python: symbol lookup error: /home/douglass/anaconda3/envs/homogenizer/lib/python3.5/site-packages/numexpr/../../../libmkl_vml_mc3.so: undefined symbol: mkl_serv_getenv

A conda update -all in both my root conda environment and my homogenizer environment fixed the problem like @ozancaglayan suggested above.

[endrebak]

The previous solution did not work for me. I had to

conda install mkl-rt

to avoid those errors.

[kapilsh]

Same as @endrebak for me. I had to reinstall mkl-rt.

[endrebak]

I got an Intel MKL FATAL ERROR: Cannot load libmkl_avx.so or libmkl_def.so sometimes after this fix, but using

conda install --yes mkl mkl-service

got rid of those too. See ContinuumIO/anaconda-issues#698

[github-actions]

Hi there, thank you for your contribution to Conda!

This issue has been automatically locked since it has not had recent activity after it was closed.

Please open a new issue if needed.