Package 'slurmtools'

throws error if the number of requested CPUs exceeds the number of CPUs available in the requested partition

Description

throws error if the number of requested CPUs exceeds the number of CPUs available in the requested partition

Usage

check_slurm_partitions(ncpu, partition, cache = TRUE)

Arguments

ncpu	number of CPUs requested by user
partition	name of partition requested by user
cache	optional argument to forgo caching

Examples

## Not run: 
check_slurm_partitions(17, "cpu2mem4gb")
check_slurm_partitions(3, "cpu2mem4gb")
check_slurm_partitions(5, "cpu4mem32gb")
check_slurm_partitions(5, "cpu4mem32gb")
check_slurm_partitions(100, "cpu32mem128gb")
check_slurm_partitions(2, "cpu2mem4gb")

## End(Not run)

---

get slurm jobs

Description

get slurm jobs

Usage

get_slurm_jobs(user = NULL)

Arguments

user	user filter

---

get list of partition names for the given cluster

Description

get list of partition names for the given cluster

Usage

get_slurm_partitions(cache = TRUE)

Arguments

cache

optional argument to forgo caching

---

get table of each partition's number of CPUs and memory

Description

• gets the raw string output from run_sinfo()

• converts to a data frame

• reorders and removes * from default partition with process_slurm_partitions()

Usage

lookup_partitions_by_cpu(cache = TRUE)

Arguments

cache

optional argument to forgo caching

Value

the processed table of each partition's number of CPUs and memory

---

get partition suggestions

Description

In a call to submit_slurm_model(), if the number of requested CPUs exceeds the number of CPUs available in the requested partition, check_slurm_partitions() errors.
This function follows up with a message providing one or two suggestions for the partition with the smallest sufficient number of CPUs and least amount of memory.
If there are no partitions with enough CPUs to accommodate the number requested, this function's return message clarifies this.

Usage

partition_advice(ncpu, partition, avail_cpus_table, cache)

Arguments

ncpu	number of CPUs requested by user
partition	name of partition requested by user
avail_cpus_table	table of partitions with respective number of CPUs and memory
cache	optional argument to forgo caching

Value

string with suggestion upon check_slurm_partitions() error

---

manipulate partition table for usability

Description

manipulate partition table for usability

Usage

process_slurm_partitions(table)

Arguments

table

the table created by lookup_partitions_by_cpu

Value

the table such that the default partition will be first and will have the asterisk removed

---

get list of each partition's number of CPUs and memory

Description

get list of each partition's number of CPUs and memory

Usage

run_sinfo()

Value

the raw partition-cpu-memory string output from sinfo

---

submit a nonmem model to slurm in parallel

Description

submit a nonmem model to slurm in parallel

Usage

submit_nonmem_model(
  .mod,
  partition = get_slurm_partitions(),
  ncpu = 1,
  overwrite = FALSE,
  dry_run = FALSE,
  ...,
  slurm_job_template_path = getOption("slurmtools.slurm_job_template_path"),
  submission_root = getOption("slurmtools.submission_root"),
  bbi_config_path = getOption("slurmtools.bbi_config_path"),
  slurm_template_opts = list()
)

Arguments

.mod	a path to a model or a bbi nonmem model object
partition	name of the partition to submit the model
ncpu	number of cpus to run the model against
overwrite	whether to overwrite existing model results
dry_run	return the command that would have been invoked, without invoking
...	arguments to pass to processx::run
slurm_job_template_path	path to slurm job template
submission_root	directory to track job submission scripts and output
bbi_config_path	path to bbi.yaml file for bbi configuration
slurm_template_opts	choose slurm template

---