Illegal Instruction on on install of qiime2-2022.8

Hi there. I'm getting an illegal instruction error when Qiime2 tries to cache my current deployment. I'm running qiime in a new conda environment on a Ubuntu 20.04 machine. My conda version is 4.14.0

I am following the installation instructions to the letter:

conda update conda
conda env create -n qiime2-2022.8 --file qiime2-2022.8-py38-linux-conda.yml
conda activate qiime2-2022.8
qiime --help

This yields the error:

QIIME is caching your current deployment for improved performance. This may take a few moments and should only happen once per deployment.
Illegal instruction (core dumped)

I'm not sure what's going wrong here. I have successfully installed older versions of Qiime on this computer before. For example, this works just fine:

conda env create -n qiime2-2022.2 --file qiime2-2022.2-py38-linux-conda.yml
conda activate qiime2-2022.2
qiime --help

Hi @alexkrohn,

Thanks for reaching out!

This is certainly an interesting error - looks like we've come across this a couple of times before with folks on Ubuntu using virtual machines to run QIIME 2:

A couple of initial questions for you:

  1. Are you doing a native install on Ubuntu, or using a virtual machine?
  2. Do you have numpy installed separately in the environment you are attempting to install QIIME 2 within? And if so, what version of numpy is it? You can run conda info to list all of the current packages/versions in your environment.

Cheers :lizard:

Hi @lizgehret!

I am running it on a native install of Ubuntu, not a virtual machine. My apologies if I missed this error before, but I didn't see any problems with native installs in the forum.

In the new qiime environment where qiime does not work (qiime2-2022.8), the numpy version is 1.23.2. In my base environment, where qiime is not installed, numpy is not installed.

conda activate qiime2-2022.8
conda list | grep numpy
# numpy                     1.23.2           py38h3a7f9d9_0    conda-forge
conda deactivate
conda list | grep numpy

Hey @alexkrohn,

Sorry you're running into this, Ubuntu is certainly supported, and a common platform for us. I expect there's something misbehaving between what your processor supports, and what math library is being used (such as openBLAS or MKL, which numpy uses).

I guess to get started, could you tell us more about your processor? Ideally we would like to see the info in /proc/cpuinfo, so if you could run this command and upload the result, that would help out:

cat /proc/cpuinfo > cpuinfo.txt

Next could you provide the result of (in your environment):

conda list blas


conda list mkl

(I expect this one to be empty).

Last step, could you run this small script which will load each QIIME 2 plugin and print each import involved? Then if you could upload the result, I expect the last line to be our culprit:

python > more_info.txt  # remember to download the script
1 Like