snakemake-storage-plugin-rucio 0.4.3

A Snakemake storage plugin that handles files available through Rucio.

snakemake-storage-plugin-rucio

A Snakemake storage plugin that handles files available through Rucio.

Usage

A manual for using different storage providers with Snakemake is available here. Documentation for this plugin is available here. Below are some examples of using the plugin in a Snakemake rule.

Download files

Download all input files and then run the workflow.

Snakefile content:

rule download:
    input:
        storage("rucio://test_scope/test{sample}.txt")
    output:
        "results/test{sample}.txt"
    shell:
        "mv {input} {output}"

This command will download files test1.txt and test2.txt from scope test_scope and move them to results/test1.txt and results/test2.txt respectively. The --verbose flag is useful to print debug logging information in case things do not work on the first attempt.

snakemake --cores 2 --verbose results/test1.txt results/test2.txt

Only get the URLs for later download

This is useful if the workflow processes multiple input files on multiple CPU cores and you would like to overlap download with computations, or if there is not enough storage space available to download all files prior to processing them.

Snakefile content:

rule get_url:
    input:
        storage("rucio://testing/test{sample}.txt", retrieve=False)
    output:
        "results/url{sample}.txt"
    shell:
        "echo {input} > {output}"

This command will store the URLs to files test1.txt and test2.txt from scope test_scope in files results/url1.txt and results/url2.txt respectively. In a real workflow, these URLs would be used to download the file when it is needed.

snakemake --cores 2 results/url1.txt results/url2.txt

Only get the URL and stream the data

This is useful if your input files are large and you only need part of the data or the data does not fit in local storage.

Snakefile content:

rule stream_file:
    input:
        storage("rucio://test_scope/test{sample}.txt", retrieve=False)
    output:
        "results/stream{sample}.txt"
    run:
        # Stream the file content into the output file.
        import gfal2
        import sys
        from pathlib import Path
        input_url  = input[0]
        output_path = output[0]
        print(f"Copying from {input_url} to {output_path}")
        Path(output_path).parent.mkdir(parents=True, exist_ok=True)

        ctx = gfal2.creat_context()
        size = ctx.stat(input_url).st_size
        file = ctx.open(input_url, "r")
        chunk_size = 2 # read 2 byte chunks for demonstration purposes
        n_chunks = (size // chunk_size) + 1
        with open(output_path, "w") as out_file:
            for _ in range(0, n_chunks):
                data = file.read(chunk_size)
                out_file.write(data)

This command retrieves the URLs to files test1.txt and test2.txt from scope test_scope and streams their content in 2 byte chunks to files results/stream1.txt and results/stream2.txt respectively. In a real workflow, larger chunks or a smarter access pattern that only reads the required bits are recommended.

snakemake --cores 2 --verbose results/stream1.txt results/stream2.txt

Upload a file

Upload a file using Rucio.

Snakefile content:

rule upload:
    output:
        "rucio://test_scope/test_file.txt"
    message:
        "Writing Hello world to {output} and uploading"
    shell:
        """
        echo "Hello world" > {output}
        """

This command will write some text to a local file test_file.txt and upload it to Rucio. The file will be uploaded to a storage element matching the RSE expression TEST_RSE_EXPRESSION in the scope test_scope and attached to the dataset test_dataset. Specifying the target dataset is required to avoid creating a replication rule per file, which would make the number of replication rules unmanageable.

snakemake --default-storage-provider rucio --storage-rucio-upload-rse TEST_RSE_EXPRESSION --storage-rucio-upload-dataset test_dataset --cores 1 --verbose 'rucio://test_scope/test_file.txt'

Contributing

Contributions are very welcome. Instructions on how to get started are available in the contribution guidelines.