container.def

#!/bin/bash

type=$1
num=$2
datafile=$3

$type -n $num $datafile > $datafile.sub
[christina.koch@ap40 automation]$ cat simple.def 
Bootstrap: docker
From: hub.opensciencegrid.org/htc/rocky:9

%files
  extract.sh /opt

%environment
  export PATH=/opt/:$PATH

extract.py

import htcondor
import classad

input_list = ["SRR2584866_2", "SRR2589044_1", "SRR2589044_2", "SRR2584866_1", "SRR2584863_1", "SRR2584863_2"]

itemdata = []
for samplename in input_list:
    itemdata.append({"sample": samplename})
function = 'head'
num = 5

upload_bucket = "osdf:///ospool/ap40/data/christina.koch/output-buckets"
singularity_image = "osdf:///ospool/ap40/data/christina.koch/singularity_imgs/extract.sif"
input_bucket = "osdf:///ospool/uc-shared/public/osg-training/tutorial-fastqc/data"

job = htcondor.Submit({
    "container_image": singularity_image,
    "universe": "container",
    "executable": "/opt/extract.sh",
    "transfer_executable": "False",
    "should_transfer_files" : "True",
    "transfer_input_files" : f"{input_bucket}/$(sample).trim.sub.fastq",
    "arguments": f"{function} {num} $(sample).trim.sub.fastq",
    "transfer_output_remaps": f'"$(sample).trim.sub.fastq.sub={upload_bucket}/$(ClusterID)/$(sample).trim.sub.fastqc.sub"',
    "output": "logs/extract-$(ProcId).out",  # output and error for each job, using the $(ProcId) macro
    "error": "logs/extract-$(ProcId).err",
    "log": "logs/extract.log",               # we still send all of the HTCondor logs for every job to the same file (not split up!)
    "request_cpus": "1",
    "request_memory": "1GB",
    "request_disk": "1GB",
})

schedd = htcondor.Schedd()

submit_result = schedd.submit(job, itemdata = iter(itemdata))  # submit one job for each item in the itemdata

extract.sh

#!/bin/bash

type=$1
num=$2
datafile=$3

$type -n $num $datafile > $datafile.sub

container.def and extract.sh are used to build a container, which can be added to the OSDF.

extract.py takes a list of input data and submits a job (using the container/script) for each item, putting the results back into an OSDF location.

slack transcript:

• can you use the python bindings to submit jobs to a remote schedd, without having a full htcondor install on your local machine?
• https://htcondor.readthedocs.io/en/latest/apis/python-bindings/api/version2/htcondor2/schedd.html#htcondor2.Schedd
• You would generally write something like

collector = htcondor2.Collector("remote-pool-cm.host.tld")
location = collector.locate(htcondor2.DaemonType.Schedd, "name-of-schedd-if-more-than-one")
schedd = htcondor2.Schedd(location)

• You do not need anything installed beyond the bindings, but you may need a Condor config file to set authentication-related settings (or set them directly in your python script)
• (... I don't remember what locate() returns if it fails to find anything; if it's None you may need to check for that explicitly to avoid accidentally submitting/trying to submit to the local schedd, instead.)
• But if you have an IDToken in the usual location, that will work without any config files.
• …or do the bindings complain if the default config file doesn’t exist?
• The bindings may well complain, but I think they work anyway.
• They spit out a warning if they don't see a config file
• A blank file in the right spot or even setting the environment to look at /dev/null is enough to get it to not emit that warning (edited)