The main reason for this behaviour is: Your links that were created in step 1 did run out of time. Process fetch_encrypted_s3_url gets pre-authenticated S3 URLs for the requested BAM files and passes these to process featureCounts. If you have many BAM files that in total take more than a day for downloading and processing (> ~30), you might want to split the workload to multiple machines. More detailed information can be found here.
The reason for this is mainly process fetch_encrypted_s3_urlthat relies on the Overture Score-client for fetching pre-authenticated URLs from AWS S3. While making a Bioconda or conda-forge recipe for this tool available wouldn't be a general issue, the requirement to use Oracle Java for running causes some trouble. Neither Bioconda nor conda-forge provide access to Oracle Java (and instead rely on OpenJDK/OpenJRE), thus making packaging the score client unfeasible for now.
Please make sure to specify a manifest file in the correct format. These are automatically created by the ICGC/DCC Data Portal in the correct .tsv.gz format and should only be unpacked to tsv format. Do not modify the files as the header is required to read the object_id and file_name information correctly. Further requirements:
- The path must be enclosed in quotes (
'or")
If the pipeline can't find your manifest then you will get the following error
ERROR ~ Manifest file not found
If you still have an issue with running the pipeline then feel free to contact us. Have look at the pipeline website to find out how.
If you have problems that are related to Nextflow and not our pipeline then check out the Nextflow gitter channel or the google group.