Hi team! we're running nf-core pipelines on AWS HealthOmics and seeing a startup failure related to

igenomes_base

. This affects several pipeline that uses nf-schema v2.x (e.g. nf-core oncoanalyser and nf-core rnaseq ), but not older workflow that uses the older nf-validation v1.x (e.g. nf-core methylseq). The actual error is

ERROR ~ Failed to read S3 object attributes for /ngi-igenomes/igenomes/

. We use our own local copy of the references so we don't need this path at all. We went through a few troubleshooting steps: 1. Set`igenomes_ignore = true`— fails 2. Set`igenomes_base = null`in config — still fails 3. Removed`"format": "directory-path"`from`igenomes_base`in`nextflow_schema.json`— works Our theory is that nf-schema v2.x validates the

igenomes_base

S3 path for accessibility regardless of

igenomes_ignore

, and this check fails because HealthOmics can't reach that S3 bucket. Not sure if this is a HealthOmics networking issue or an nf-schema issue — are we missing something, or is this a known limitation with directory-path validation on restricted environments?

No idea, but I think S3 paths are more tricky to validate, because they need more layers of tools than a local file. There is a discussion regarding this here. Besides editing the schema, you could also explicitly ignore

igenomes_base

in the validation (see link). Maybe @Nicolas Vannieuwkerke has better ideas.

Thanks Thomas. For some reasons setting it to ignore didn't work. However, just found out that setting the path to null in nextflow.config works.

Hi yes this is a bug in the pipelines and not really an nf-schema bug. I've been advocating for pipelines to remove validation in

--igenomes_base

since it has an S3 path as default. The validation should only be done on the actual paths derived from that

Thanks Nicolas, you have solved the mystery!

All of them 😅. I just fixed this in the template

was fixed in rnaseq a couple of releases ago