Hey! In case you haven't seen it, the full size release test failed: https://cloud.seqera.io/orgs/nf-core/workspaces/AWSmegatests/watch/S4FTJJjJfkdbZ Seems to be something with unicycler and fusion @Florian Wuennemann you looked into that recently, right?

Hi @Rike Hanssen, do you mean that the

-profile test_full

failed? I just ran it and it worked for me (although it took a long time to finish).

nextflow run nf-core/bacass -r dev \
    -profile test_full,docker \
    --outdir results \
    -resume

can you run it with the tagged release that is used in the linked run?

Thanks for pinging me @Rike Hanssen! I did have issues with unicycler in nf-core/viralrecon as well. I believe the failing Megatests might be due to the interaction of Unicycler and FUSION, this was what caused the problem in viralrecon. I am not 100% sure this is the same error yet, but will test the bacass test on our end with the fix to see if that works! The temporary fix for viralrecon when running with FUSION was to specify an exception for unicycler:

withName:UNICYCLER {
        scratch = true
    }

@Simon Pearce and I also just updated the unicycler module to the newest version, which might also fix this issue: https://github.com/nf-core/modules/pull/5544#event-13205731695 So I see 3 options: 1) we add the exception to the Megatests to make it pass (but it will still fail for people using this configuration with FUSION on their end), 2) We introduce the fix into the config when using unicycler 3) We check whether updating the unicycler module fixes this issue and make a quick patch for bacass and viralrecon

can you run it with the tagged release that is used in the linked run?

Sorry, I cannot access the URL. It says "Organization with the name 'nf-core' not found". Which version of nf-core/bacass are you using?

@Daniel_VM This is the git revision commit that @Rike Hanssen mentioned:

d230ebf3d77e030bdcd0d002ef082a3870c68d99

which is version 2.3.0

Awesome, running this revision now. Thanks. I am going to open an issue to update

local/unicycler

to

nf-core/unicycler

to avoid potential versioning issues, as stated in the PR. 😉

Quick update from my test. I relaunched the Megatest adding:

process {
    withName:UNICYCLER {
        scratch = true
    }
}

To make unicycler use scratch with FUSION. The execution log for Unicycler didn't load and the run was still "running" after 53 minutes, so I canceled it, thinking something was up. After cancelling, the unicycler logs started appearing, so it seems that with scratch it just really takes that long to run unicycler on this test data? 🤷

Sure. I was surprised that the

nf-core/bacass

test_full

profile took so long. Once I finish running version 2.3.0 (

-r 2.3.0

), I'll let you know the run times. 👍

I can confirm that the fix for FUSION with

scratch = true

fixes the AWS Megatest issue @Rike Hanssen https://cloud.seqera.io/orgs/nf-core/workspaces/AWSmegatests/watch/hcogCMTmp0NIj The total runtime with this fix for Megatests was 1h 2 m 4s!

ok great, thanks @Florian Wuennemann. So then things are fixed for now and the updated unicycler should fix it for the next release automatically, right?

I will have to test with the new unicycler and FUSION to make sure that fixes it. If the new version doesn't fix it, we would have to build in the scratch = true for the unicycler process exclusively to fix it for FUSION based runs.

Hii!

-profile test_full

took 5h to complete 🥴 . I attach the execution timeline. Notably, Unicycler (sample SRR10093029) alone took 1h 37min.

I am not particularly concerned about the time. These tests are there to mimic a real life scenario.

Great 👍