Issue with server/headless environments

[dkoslicki]

Running the code locally on a system with a GUI is no problem. But when trying to run on a server (running the same version of Ubuntu, but with no GUI), the following error is encountered:

$ python src/profile_to_plot.py -i data/prediction_multi.profile -g data/ground_truth_multi.profile -b test -s sample_0 genus
qt.qpa.xcb: could not connect to display
qt.qpa.plugin: Could not load the Qt platform plugin "xcb" in "" even though it was found.
This application failed to start because no Qt platform plugin could be initialized. Reinstalling the application may fix this problem.

Available platform plugins are: eglfs, minimal, minimalegl, offscreen, vnc, webgl, xcb.

Aborted (core dumped)

Much googling lead nowhere.

[luizirber]

This sounds like an issue with matplotlib defaulting to qt, you can try to add a matplotlibrc file like we do in sourmash: https://github.com/dib-lab/sourmash/blob/master/matplotlibrc

For conda it also helps to use matplotlib-base instead of matplotlib, I think it avoids bringing qt and all the Xorg libs (which also makes environments way smaller)

[dkoslicki]

Thanks for the suggestion @luizirber! I’ll give that a try (though I think ete3 brings in qt as well, so might be two issues to address).

[dkoslicki]

Turns out just adding os.environ['QT_QPA_PLATFORM']='offscreen' before importing matplotlib, seaborn, or ete3 did the trick. So no need to switch to matplotlib-base (though it may indeed make the env smaller, so may be worth doing)