kmetashot works in conda not in docker

[paulzierep]

After a long time of digging, I realized this tool does not work in the current biocontainer.
This was tested on a 60 GB memory machine.
Conda test:
tool_test_output-conda.html
Container test:
tool_test_output.html

I tested the tool itself in the container:

docker run --rm -it \
  -v /vol/data/galaxytools/tools/kmetashot/test-data:/data:ro \
  quay.io/biocontainers/kmetashot:2.0--pyh7e72e81_1 \
  bash

root@069cd56dd454:/data# kMetaShot_classifier_NV.py \
-b bins \
-o output \
-r kMetaShot_reference.h5 \
-p 1 \
-a 0 

                           ################################################
                           #        kMetaShot Classifier Algorithm        #
                           #                  Version 2.0                 #
                           #               Defazio G. et al.              #
                           ################################################

[2026-01-14 09:35:05] Start Assignment
Namespace(bins_dir='bins', reference='kMetaShot_reference.h5', ass2ref=0.0, processes=1, out_dir='output')
The output/bins already exists
[2026-01-14 09:35:07] Reference loading ...
Bus error (core dumped)
root@069cd56dd454:/data# /usr/local/lib/python3.13/multiprocessing/resource_tracker.py:324: UserWarning: resource_tracker: There appear to be 1 leaked shared_memory objects to clean up at shutdown: {'/psm_05a2fcac'}
  warnings.warn(

In the conda env it works.

[paulzierep]

ping @SantaMcCloud, @gdefazio

[gdefazio]

The problem may be the dimension of /dev/shm in the container, because after reference loading this object becomes a shared object.

[gdefazio]

@paulzierep
I tried to gain the space of /dev/shm up to 20 gb as follows:

docker run --rm -it --shm-size=20g  \
-v /vol/data/galaxytools/tools/kmetashot/test-data:/data   \
quay.io/biocontainers/kmetashot:2.0--pyh7e72e81_1   bash

then I tried to run kMetaShot and it works.

-b bins \
-o output1 \
-r kMetaShot_reference.h5 \
-p 1 \
-a 0 

                           ################################################
                           #        kMetaShot Classifier Algorithm        #
                           #                  Version 2.0                 #
                           #               Defazio G. et al.              #
                           ################################################

[2026-01-20 13:41:42] Start Assignment
Namespace(bins_dir='bins', reference='kMetaShot_reference.h5', ass2ref=0.0, processes=1, out_dir='output1')
[2026-01-20 13:41:44] Reference loading ...
[2026-01-20 13:42:04] Reference loading DONE
[2026-01-20 13:42:04] START bins/all_contig.fasta.gz 
[2026-01-20 13:42:31] STOP bins/all_contig.fasta.gz
[2026-01-20 13:42:36] Assignment of bins sequences DONE

[paulzierep]

Nice ! Now the question is how to fix this. I assume there is Galaxy way @bgruening, do you have any idea, can we (a) only run this in conde env and set TPV rule to 40 gb or run in docker and set --shm-size=20g ?
Maybe the tool could be changed to require less resources, probably with the drawback to longer runtime.
But if it can be fixed on our side it is most likely more simple.

[bgruening]

We can configure Docker do run with this option. No problem.

But the TS version of that tool is currently broken:

kmetashot.xml - This file requires an entry in the tool_data_table_conf.xml file. Upload a file named tool_data_table_conf.xml.sample to the repository that includes the required entry to correct this error.