kuberun does not export function defined in project nextflow.config

[micans]

Bug report

I run the exact same pipeline on LSF and k8s (almost identical command lines,
using kuberun for the latter):

nextflow run cellgeni/rnaseq-noqc -profile farm3 --samplefile $samplefile --genome GRCh38 --runtag $sid --studyid $sid
nextflow kuberun cellgeni/rnaseq-noqc -profile docker --samplefile $samplefile --genome GRCh38 --runtag $sid --studyid $sid -c nf.config -v nf-pvc:/mnt/gluster/

Expected behavior and actual behavior

On LSF this works fine. I expect it to work in the same way on k8s. The first process runs fine on k8s. The second process uses a function check_max defined in the project nextflow.config (https://raw.githubusercontent.com/cellgeni/rnaseq-noqc/master/nextflow.config). The first instance of the second process fails immediately with the message below:

Program output

(many irods processes)
. . .
[c9/4ee3f1] Submitted process > irods (febc_C)
[57/f08b79] Submitted process > irods (febc_D)
ERROR ~ Error executing process > 'crams_to_fastq (coxy_A)'

Caused by:
  No signature of method: check_max() is applicable for argument types: (nextflow.util.MemoryUnit, java.lang.Stri
Possible solutions: check_log(java.lang.Object)

Steps to reproduce the problem

I'll describe more context here. The nextflow log file is attached. The nf.config file from kuberun is this:

params { genomes_base = "/mnt/gluster/data/nf-inputs" }
executor { queueSize = 30 }
k8s { userDir = '/mnt/gluster/svd/tic-97' }

process {
  maxForks = 30
  afterScript = 'sleep 2'
}

After execution, if I visit k8s.userDir, there is a nextflow.config file there created by NF itself, clearly representing a dump of internal configuration state or a subset thereof. In this file check_max is absent. It seems as if the function definition is not exported. I can make a reproducible test-case, but it will require more time and I have some hope that this describes the problem to a good extent.

Environment

I've tested this both with 32 (version 0.32.0 build 4897) and with 31 (don't have version ready):

bc-2,TIC-misc/tic-97, nextflow -version

      N E X T F L O W
      version 0.32.0 build 4897
      last modified 27-09-2018 10:17 UTC (11:17 BST)
      cite doi:10.1038/nbt.3820
      http://nextflow.io

bc-2,TIC-misc/tic-97, java -version
java version "1.8.0_74"
Java(TM) SE Runtime Environment (build 1.8.0_74-b02)
Java HotSpot(TM) 64-Bit Server VM (build 25.74-b02, mixed mode)
bc-2,TIC-misc/tic-97, uname -a
Linux bc-29-3-02 3.2.0-105-generic #146-Ubuntu SMP Fri Jun 10 20:10:44 UTC 2016 x86_64 x86_64 x86_64 GNU/Linux

nextflow.log

[pditommaso]

The problem is that kuberun creates its own config file and was not taken in consideration to possibility of defining custom function in the NF configuration.

A solution could be provided implementing check_max as a built-in function as suggested by #640.

In any case I need to verify if it's possible to support custom method in nextflow.config file when using kuberun command.

[pditommaso]

The idea to solve this issue is to move custom functions defined in the config file in library files. This requires #940.

[courcelm]

I also ran into the check_max issue while running the nf-core/mhcquant on Kubernetes. This issue is almost one year old. Any plan to fix this?

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.