Error in executing PROKKA

[anugos]

Error in executing PROKKA

[jfy133]

Hi @anugos
This seems to be a common and 'unresolved' prokka error.
The recommendation is posted here: tseemann/prokka#402 (comment)

Please install PROKKA manually (e.g. via conda), cd into the work directory reported into the error, then use the command in the .command.sh file to re-run prokka, but without redirecting the stdout/in

[roberta-davidson]

Hey @anugos @jfy133 ! Found a bit of a workaround. I downloaded this container for Prokka and then modified my config to use this container. Also would not work via slurm submssion to our HPC, but did on the head node (?!), and then had to modify to run 1 at a time so tmp directories for prokka didn't overwrite eachother. On second thought, maybe using a different container was unnecessary but anyway..
Overall additions to config file:

process {
   executor = 'slurm'
   clusterOptions="-N 1 -p skylake,icelake"
  withName: PROKKA {
    container = '/<path>/prokka_1.14.6--pl5321hdfd78af_5.sif'
    executor = 'local'
    maxForks = 1
  }
}

[jfy133]

@roberta-davidson huh interesting... what was the actual error for you (i.e., what was otehrwise piped to nothing?

Is it a /tmp clash or something? This we can maybe set to use a the process' specific work directory...

[roberta-davidson]

The original command in error from running mag was:

  [17:36:49] Could not run command: cat MEGAHIT\-MetaBAT2\-Bushfire_A24936\.15\/MEGAHIT\-MetaBAT2\-Bushfire_A24936\.15\.IS\.tmp\.44\.faa | parallel --gnu --plain -j 2 --block 43333 --recstart '>' --pipe blastp -query - -db /usr/local/db/kingdom/Bacteria/IS -evalue 1e-30 -qcov_hsp_perc 90 -num_threads 1 -num_descriptions 1 -num_alignments 1 -seg no > MEGAHIT\-MetaBAT2\-Bushfire_A24936\.15\/MEGAHIT\-MetaBAT2\-Bushfire_A24936\.15\.IS\.tmp\.44\.blast 2> /dev/null

I really don't understand why this workaround works...
I set out to do as you suggest above, and wrote a script to run .command.sh in each work dir using my own prokka container, and then @shyama-mama figured out to just adjust the config and point to that container when the pipeline runs. Then realised that .command.sh with my container worked on head node but not within the pipeline (no idea why). Then adjusted to execute locally, and one at a time.