nf-core/configs: OIST Configuration

To use, run the pipeline with -profile oist. This will download and launch the oist.config, which has been pre-configured for the clusters of the Okinawa Institute of Science and Technology Graduate University (OIST). Using this profile, container images are pulled and converted to Singularity images before execution of the pipeline.

The oist profile is always required. On its own, it submits tasks to the compute partition of Deigo, on the AMD nodes (-C epyc), with at most 128 CPUs, 500 GB and 96 hours per task. To run elsewhere, add one of the sub-profiles after it.

Loading Nextflow

ml purge
ml bioinfo-ugrp-modules Nextflow3

Pipelines released before they were made compatible with the Nextflow strict syntax fail with Nextflow3, typically with a syntax error in nextflow.config or with a boolean parameter reported as a string. Load Nextflow2 instead of Nextflow3 to run them.

Work directory

By default, Nextflow creates its work directory in the directory where it is launched. When launching from anywhere other than /flash on Deigo or /work on Saion, for example from /bucket or /home, always give the work directory with -w:

nextflow run <pipeline> -profile oist -w /flash/<unit>/<user>/work_<run> ...

/bucket is read-only on compute nodes, and a work directory there makes the run hang and then fail with an uninformative error.

Interaction with Slurm

Job arrays are incompatible with some pipelines:

  • Arrays are not supported by the local executor. A pipeline that runs some processes locally, or a run where all tasks use the local executor, stops with Executor 'local' does not support job arrays.
  • All tasks of an array share the same resources. Pipelines that give different resources to each task of a process can hang.

To disable arrays, add -process.array 0 to the command line. As of September 2026, the pipelines of sanger-tol and nf-core/genomeassembler need it.

Sub-profiles

A sub-profile is added after oist, for example -profile oist,oist_short, and replaces only the settings below.

Sub-profile Cluster Partition
oist_short Deigo short
oist_saion Saion intel
oist_ci Deigo short

oist_saion lets Slurm requeue preempted jobs instead of cancelling them, and waits up to one day for a requeued job to restart before Nextflow considers the task failed.

oist_ci is for test suites of tens of short tasks, for example -profile oist,test,oist_ci. Its limits replace those of the pipeline’s test profile. It also sets relaxed falues for submitRateLimit, pollInterval and queueStatInterval.

Config file

See config file on GitHub

conf/oist
// nf-core/configs: OIST Deigo cluster profile
params {
config_profile_description = 'The Okinawa Institute of Science and Technology Graduate University (OIST) HPC cluster profile provided by nf-core/configs.'
config_profile_contact = 'OISTs Bioinformatics User Group <BioinfoUgrp@oist.jp>'
config_profile_url = 'https://github.com/nf-core/configs/blob/master/docs/oist.md'
// Legacy limits, read only by pipelines made with nf-core/tools < 3.0
max_memory = 500.GB
max_cpus = 128
max_time = 96.h
}
validation {
ignoreParams = ['max_memory', 'max_cpus', 'max_time']
}
singularity {
enabled = true
}
process {
resourceLimits = { return [
memory: params.max_memory,
cpus: params.max_cpus,
time: params.max_time,
]}
executor = 'slurm'
queue = 'compute'
clusterOptions = '-C epyc'
// Must be 0 with the local executor, and not larger than executor.queueSize.
array = 10
}
executor {
// Throttle the Slurm controller. Form is 'count/interval'.
submitRateLimit = '12/1min'
queueSize = 208
// Checks .exitcode files on /flash; does not query Slurm.
pollInterval = '30 sec'
// Period of the only recurring squeue call.
queueStatInterval = '5 min'
}
profiles {
// Deigo, 'short' partition on the Intel nodes: -profile oist,oist_short
// queueSize and array are those of 'compute'.
oist_short {
params {
max_cpus = 40
max_time = 2.h
}
process {
queue = 'short'
}
}
// Saion, 'intel' partition: -profile oist,oist_saion
oist_saion {
params {
max_memory = 120.GB
max_cpus = 40
}
process {
queue = 'intel'
// Preempted jobs are requeued by Slurm instead of cancelled.
clusterOptions = '--requeue'
array = 10
}
executor {
queueSize = 40
// A requeued job is pending, which Nextflow counts as not running;
// wait for it to restart before declaring the task failed.
exitReadTimeout = '1 day'
}
}
// Test suites (tens of short tasks) on Deigo: -profile oist,test,oist_ci
// Limits are those of the nf-core template's test profile, which the
// institutional profile would otherwise replace.
oist_ci {
params {
max_memory = 15.GB
max_cpus = 4
max_time = 1.h
}
process {
queue = 'short'
array = 0
}
executor {
queueSize = 32
submitRateLimit = '1/1s'
pollInterval = '5 sec'
queueStatInterval = '1 min'
}
}
}