Testing the bacass pipeline with -profile test,doc...
# bacass
r
Testing the bacass pipeline with -profile test,docker but still getting an error when running tbl2asn even though I changed prokka to 1.14.6 in the environment.yml file. I get "Could not run command: tbl2asn -V b -a r0k...." @Curtis Kapsak @Till Bayer how did you get it to work?
c
I would double check that you’re using the right version of the prokka conda env or biocontainer if using docker. @Till Bayer could you share the specific version that worked for you? As a workaround, I have a prokka docker image that should work. Search staphb/prokka on dockerhub to find the repo
r
Thanks @Curtis Kapsak! I thought this was using a mix of conda and docker. So i thought changing the environment.yml to reflect the newest prokka version would have fixed it. Where do I check which biocontainer is being used since i am using docker?
c
TBH I’m not sure since I don’t use nextflow or bacass much, just lurk here 🤓. I think the docker image name and tag will be specified in a config file somewhere, just not sure where. The name will likely have ‘quay.io’ in it somewhere since that’s where the bio containers are hosted
r
Thanks @Curtis Kapsak I'll poke around and see what i can find
c
No problem, good luck!
a
I’ll sit down these days and release a new container with new nfcore changes
And a permanent fix for that issue
r
Sorry @Alex Peltzer to add to your plate! Meanwhile do you know of a quick fix? I'll keep trying different things meanwhile
p
The docker container is built using the conda environment file, so you’re kind of right. However the pipeline pulls the docker image when you run the pipeline instead of building it each time, so changing the environment file won’t affect the image if that makes sense.
If you changed the environment file then you can rebuild the docker container on your machine if you want, but you’ll need to know the exact name of the container that is referenced in the pipeline
Might be easier to use run with conda?
-profile test,conda
Instead of docker
t
@Ryan I used the latest biocontainer for the prokka process, that works.
a
The plan is to add the latest prokka versino that includes a version of tbl2asn that doesn't self-destroy after 6 months into the container
Thats pretty much the same as Till did here - using a different container 😉
t
Yepp
Some other processes like dfast and medaka use biocontainers, is that because they don't work with conda or something, or will they be added to the main container also?
a
Incompatible software dependencies, e.g. using python2 so we can't have them in the same "main" container
t
You could have two conda envs in one container 😉
a
We had mixed experiences with this and sometimes odd side-effects
p
Yeah, you basically can’t because we don’t use
conda activate
as you would normally
t
You would basically have to make conda activate part of the process script? Sounds messy indeed.
p
exactly, which is why we haven’t gone that way (would break for all non-docker users)
But we’re moving towards DSL2, when we will start having separate conda envs / containers for each process. So then this problem should go away..
a
Yeah - I'm staying away from that for now until DSL2 modules have settled a little. Particularly as the bacass pipeline dependencies are sometimes hard to get running fine, e.g. had lots of trouble with seemingly working Unicycler recipes from bioconda here when I made the first release of this
So bugfix release will just be a minor release including newest nf-core template, and update to prokka etc so that the pipeline works fine again without any modification required. Problem only is that I cannot fully work on this during work hours ... which means it depends a bit on my free time schedule in the evening(s).
👍🏻 1
r
Thanks @Till Bayer @Alex Peltzer @Phil Ewels for explaining all this. So does this mean that in the dockerfile I change the base from 1.7 version to 1.10.2? Not sure I understand where else to change the biocontainer.
a
No
t
@Ryan In
conf/base.conf
you can add a section similar as the last one for the prokka process.
👍 2
a
I'll merge this one here: https://github.com/nf-core/bacass/pull/29 and that one: https://github.com/nf-core/bacass/pull/30 and this one: https://github.com/nf-core/bacass/pull/19 and then upgrade prokka to the version on bioconda - that should tthen hopefully do the trick entirely
👍 1
r
I tried changing to -profile conda,test but ran into issues with other software and saw some post that the conda profile was not stable.
a
it is not
And also not complete
Maybe I remove the hints to this also