`Invalid parameter partition` error when running on other slurm clusters

[Lola-W]

Hi team,

I'm trying to adapt the pipeline for our use on the TSCC cluster here at UCSD, which also uses SLURM. And I encountered a configuration error indicates that our partition is not in the allowed list:

ERROR ~ Unable to parse config file: 'template.config'
 Invalid parameter partition. Valid values: [F16, F32, F72, F72_3TB]. 

The error seems to came from a schema validation that restricts the partition parameter to a specific list ([F16, F32, F72, F72_3TB]), while our cluster uses different partition names, such as condo.

I modified the template.config :

metapipeline-DNA/config/template.config

Line 19 in 49d3b8d

partition = 'F72'

as below

params {
    partition = 'condo'
    clusterOptions = '-N 1 -t 24:00:00 -q condo -A ourAccount --output=./log/slurm-%j.out'
}

and ran with nextflow run main.nf -c template.config in an interactive console.

I'm wondering what the recommended approach is for overriding the partition schema for different SLURM clusters. I also searched the codebase to understand where the [F16, ...] list was defined but couldn't immediately locate it. Any guidance on better configuring this pipeline would be greatly appreciated. Thank you for your help!

[yashpatel6]

Hello, we're currently working on allowing partition to be more generalized. In the meantime, it can be configured in the config/schema.yaml, which defines the validation schema for parameters

[zhuchcn]

@Lola-W Here is my trick. In your config file, specify the partition as "F72".

params {
    partition = 'F72'
}

And also in the same config file, after the methods.set_up() call, add these lines:

process['withName:call_metapipeline_DNA'].cpus = 72
process['withName:call_metapipeline_DNA'].memory = 144.GB
process['withName:call_metapipeline_DNA'].queue = 'condo'

This will let the pipeline use our "F72" configuration, but tells it to submit to the condo partition. But make sure that your condo partition has at least 72 CPU.

You probably don't need the "clusterOptions" string. Each sample will likely take more than 24 hours. You also don't need the -q since it's already specified above. You also don't need the --output cuz it will mess up the pipeline output. You can keep -A.

Let us know if you have any trouble!

[Lola-W]

Thanks for your kind reply! I just changed the config file as @zhuchcn suggested and it passed smoothly. However I ran into another issue:

ERROR ~ Unable to parse config file: 'template.config'
  /hot/resource/tool-specific-input/StableLift-1.1.0/model/GRCh37-to-GRCh38/RF-model_GRCh37-to-GRCh38_gSNP_HaplotypeCaller.Rds does not exist.

I tried unselect the StableLift from the config as below:

requested_pipelines = ['generate-SQC-BAM', 'call-gSNP', 'call-mtSNV', 'call-sSNV', 'call-sSV', 'call-gSV', 'annotate-gSNP', 'calculate-mtDNA-CopyNumber']

Also I deleted the block for StableLift in pipeline_params in the config, and set required: false for StableLift defined in the config/schema.yaml, however the Rds file is still requested. I'm wondering if it's defined somewhere else, or if it's a dependency for other pipelines? Thanks!

UPDATE:
Found stablelift_models in the config/default.config, after commenting this chunck out the pipeline works well.

[yashpatel6]

Great, that's something we're working on improving as well, with parameter validation and its interplay with pipeline selection. Glad to hear it's working