https://nf-co.re logo
Join Slack
Powered by
I am not sure if this related to <@U04JKRB6UUR>’s ...
# multiqc
e
I am not sure if this related to @Lili Andersson-Li’s problem, but the latest version of MultiQC (1.30) doesn't seem to be creating any outputs for me when using 
conda
. This is my debug-mode error on the github runner: https://github.com/nf-core/proteinfamilies/actions/runs/16718396488/job/47318451095?pr=107 The inputs are the same as when on singularity/docker. I have no idea what's going on. Had the same problem with a local conda execution, so it's not the runner's fault. Anyone else had/reported something similar?
conda log in an nf-test execution:
Copy code
/// MultiQC 🔍 v1.30

            config | Loading config settings from: multiqc_config.yml
     update_config | Prepending directory to sample names
     update_config | Not cleaning sample names
       file_search | Search path: /path/to/.nf-test/tests/c8c0bfc10c1690e69dd956035ea295f2/work/04/3eb1125babc15519f555884a6c8932

    custom_content | nf-core-proteinfamilies-methods-description: Found 1 sample (html)
    custom_content | mgnifams_test_cluster_distribution: Found 7 samples (PlotType.TABLE)
    custom_content | family_metadata: Found 5 samples (PlotType.TABLE)
    custom_content | seqkit_stats_summary: Found 1 samples (PlotType.TABLE)
    custom_content | nf-core-proteinfamilies-summary: Found 1 sample (html)

     write_results | Data        : None
           multiqc | MultiQC complete
singularity log in an end-to-end run:
Copy code
/// MultiQC 🔍 v1.30

            config | Loading config settings from: multiqc_config.yml
       file_search | Search path: /path/to/proteinfamilies/work/4e/3aa258abcca9bb54a5853530c5f693

    custom_content | nf-core-proteinfamilies-methods-description: Found 1 sample (html)
    custom_content | mgnifams_test_cluster_distribution: Found 7 samples (PlotType.TABLE)
    custom_content | family_metadata: Found 5 samples (PlotType.TABLE)
    custom_content | seqkit_stats_summary: Found 1 samples (PlotType.TABLE)
    custom_content | nf-core-proteinfamilies-summary: Found 1 sample (html)
     write_results | Rendering plots. Export plots to formats png, svg, pdf is requested, so it might take a while. To disable plot export, set `export_plots: false` in config, or remove the `--export-plots` command line flag


     write_results | Data        : multiqc_data
     write_results | Report      : multiqc_report.html
     write_results | Plots       : multiqc_plots
           multiqc | MultiQC complete
l
Hi Evangelos, I can’t generate MultiQC plots on our server with singularity, but I can generate the plots in CodeSpace. @Júlia Mir also mentioned a similar issue to mine.
e
Must be a different issue 😞
l
It seems so.
e
Investigating..because even when rolling back to 1.29 I still get the same behaviour...could be something with latest Nextflow version. Will post findings here
So, the multiqc log for conda has these 2 additional lines, which singularity doesn't (are they expected?):
Copy code
update_config | Prepending directory to sample names
update_config | Not cleaning sample names
and then produces empty results:
Copy code
write_results | Data        : None
while the singularity version produces:
Copy code
write_results | Rendering plots. Export plots to formats png, svg, pdf is requested, so it might take a while. To disable plot export, set `export_plots: false` in config, or remove the `--export-plots` command line flag

write_results | Data        : multiqc_data
write_results | Report      : multiqc_report.html
write_results | Plots       : multiqc_plots
Not sure what else to test..Nextflow version doesn't seem to alter this behaviour
j
I think you are not the only one with this issue. I hit this this morning, with fresh installs of v1.25.1 from bioconda. My workaround was to install multiqc from PyPI but I think the best we can do for now is to see if other folks (bioconda maintainers e.g.) find and fix the root cause 😕
🙏 1
e
Aye, thanks! Saw that on another pipeline as well. Guess I'll wait a bit then
p
@Vlad Savelyev are you around? Can you take a look into this please?
@Evangelos Karatzas if you're able to get a reproducible example outside of nextflow / nf-test that would be most useful.
Is it just the PNG plots not being generated? Or the entire HTML report / data files?
I know that singularity has had issues with plotly rasterisation, we added functionality to gracefully skip plot export if this happens. I've seen that mentioned a few times. Not sure if it's the same here.
MRE from @Jason (Fulcrum Genomics) :
```mamba env create -f mqc
mamba install bioconda::multiqc=1.25.1```
Activate the env and run 
multqc .
j
Thanks Phil! @Vlad Savelyev, feel free to DM me if you need more info about reprex. Would love to help!
e
Nothing was getting generated; no plot folders or .html file. But only on conda
p
Ok then this is something new 🥲
Have the biocontainer images been regenerated or anything weird? If they were working before (for older versions of multiqc) and now stopped?
It'd be good to find some tests that previously passed and now failed, to compare and try to figure out what had changed
Separately if @Vlad Savelyev or anyone can repro and dig into what's broken that'd be best
e
Bit weird behaviour..1.29 was working fine on the previous pipeline release. Now when I test locally with it it doesn't. I have another in-house pipeline that works with 1.29 but not with 1.30. I know this isn't very helpful, but...
p
Could it be something to do with locally cached singularity images?
e
I guess it could..since it always happens on the github runner with 1.30 for my pipeline, I guess I could try reverting back to 1.29 and retrying there as well
It was a locally cached problem in the end. v1.29 works fine, so for my current pipeline release I rolled-back to 1.29. Seems to be a 1.30-only conda problem
j
Evangelos, Phill, Vlad: I think I'm misunderstanding something and/or something is not adding up. I can confirm that new installations of MultiQC pinned to 1.25.1 via 
conda
were causing issues for us. On a related note, clearing the "cache" in our case the conda cache, might is a good idea. I'll try that and report back.
😕 The underlying seems to have been solved upstream. @Evangelos Karatzas, I do think that we saw something weird on Monday! This will continue to puzzle me but I think this is solved now.
e
Are we waiting for an update version of nfcore multiqc then? Or will 1.30 now work?
j
Full transparency, I'm not one of the nf-core maintainers or a multiqc maintainer! tl;dr: I think you can try installing MultiQC v1.30 and I have a hunch that it will now work. What I think happened was that something in the conda-forge/bioconda dependency solve/management/tree was broken on Monday when we had issues. It was a a strange temporary issue for which we still don't know the root cause. We were both unlucky in that we both ran workflows/programs that built conda environments installing MultiQC on Monday. Whatever was causing issue for us is now fixed.
👀 1
e
Oh I see!! Will try it out tomorrow and let you know
🙌 1
p
My favourite kind of bug if true...
j
Sorry to bother you on your holiday Phil. We have logs in completely separate CI systems we work in but we can't share those 😞 --- it's been bothering a couple of us at Fulcrum! If we figure out what the heck happened I will definitely share.
🙏🏻 1
p
No worries at all. Please keep me posted, I won't hold my breath just yet..
j
Yeah... we were chatting internally and were concerned that folks in the wider bfx ecosystem will an issue when containers relying on bioconda/conda-forge start getting rebuilt 👀. Anyway...
p
It doesn't look like the docker images have been rebuilt recently at least: https://quay.io/repository/biocontainers/multiqc?tab=tags
e
You were right, clearing and re-requesting the v1.30 conda env worked :'D
🙌 1
🙌🏻 1
Not sure if the github runners hold a local version of the conda environments; if so, that should also be manually deleted then, so it can be recreated properly
Open in Slack
PreviousNext
Powered by