Submit R Calculations to a 'SLURM' Cluster

Functions that simplify submitting R scripts to a 'SLURM' cluster workload manager, in part by automating the division of embarrassingly parallel calculations across cluster nodes.

Many computing-intensive processes in R involve the repeated evaluation of a function over many items or parameter sets. These so-called embarrassingly parallel calculations can be run serially with the lapply or Map function, or in parallel on a single machine with mclapply or mcMap (from the parallel package).

The rslurm package simplifies the process of distributing this type of calculation across a computing cluster that uses the SLURM workload manager. Its main function, slurm_apply, automatically divides the computation over multiple nodes and writes the necessary submission scripts. It also includes functions to retrieve and combine the output from different nodes, as well as wrappers for common SLURM commands.

Development of this R package was supported by the National Socio-Environmental Synthesis Center (SESYNC) under funding received from the National Science Foundation DBI-1052875.

• Basic example
• Single function evaluation
• Adding auxiliary data and functions
• Configuring SLURM options
• Generating scripts for later submission
• How it works / advanced customization

Basic example

To illustrate a typical rslurm workflow, we use a simple function that takes a mean and standard deviation as parameters, generates a million normal deviates and returns the sample mean and standard deviation.

test_func <- function(par_mu, par_sd) {
    samp <- rnorm(10^6, par_mu, par_sd)
    c(s_mu = mean(samp), s_sd = sd(samp))
}

We then create a parameter data frame where each row is a parameter set and each column matches an argument of the function.

pars <- data.frame(par_mu = 1:10,
                   par_sd = seq(0.1, 1, length.out = 10))
head(pars, 3)

  par_mu par_sd
1      1    0.1
2      2    0.2
3      3    0.3

We can now pass that function and the parameters data frame to slurm_apply, specifiying the number of cluster nodes to use and the number of CPUs per node. The latter (cpus_per_node) determines how many processes will be forked on each node, as the mc.cores argument of parallel::mcMap.

library(rslurm)
sjob <- slurm_apply(test_func, pars, jobname = "test_job", 
                    nodes = 2, cpus_per_node = 2)

The output of slurm_apply is a slurm_job object that stores a few pieces of information (job name and number of nodes) needed to retrieve the job's output.

Assuming the function is run on a machine with access to the cluster, it also prints a message confirming the job has been submitted to SLURM.

Submitted batch job 352375

Particular clusters may require the specification of additional SLURM options, such as time and memory limits for the job. Also, when running R on a local machine without direct cluster access, you may want to generate scripts to be copied to the cluster and run at a later time. These topics are covered in additional sections below this basic example.

After the job has been submitted, you can call print_job_status to display its status (in queue, running or completed) or call cancel_slurm to cancel its execution. These functions are R wrappers for the SLURM command line functions squeue and scancel, respectively.

Once the job completes, get_slurm_out reads and combines the output from all nodes.

res <- get_slurm_out(sjob, outtype = "table")
head(res, 3)

      s_mu       s_sd
1 1.000005 0.09987899
2 2.000185 0.20001108
3 3.000238 0.29988789

When outtype = "table", the outputs from each function evaluation are row-bound into a single data frame; this is an appropriate format when the function returns a simple vector. The default outtype = "raw" combines the outputs into a list and can thus handle arbitrarily complex return objects.

res_raw <- get_slurm_out(sjob, outtype = "raw")
res_raw[1:3]

[[1]]
      s_mu       s_sd 
1.00000506 0.09987899 
 
[[2]]
     s_mu      s_sd 
2.0001852 0.2000111 
 
[[3]]
     s_mu      s_sd 
3.0002377 0.2998879 

The files generated by slurm_apply are saved in a folder named _rslurm_[jobname] under the current working directory.

dir("_rslurm_test_job")

[1] "params.RData"    "results_0.RData" "results_1.RData" "slurm_0.out"    
[5] "slurm_1.out"     "slurm_run.R"     "submit.sh" 

The utility function cleanup_files deletes the temporary folder for the specified slurm_job.

Single function evaluation

In addition to slurm_apply, rslurm also defines a slurm_call function, which sends a single function call to the cluster. It is analogous in syntax to the base R function do.call, accepting a function and a named list of parameters as arguments.

sjob <- slurm_call(test_func, list(par_mu = 5, par_sd = 1))

Because slurm_call involves a single process on a single node, it does not recognize the nodes and cpus_per_node arguments; otherwise, it accepts the same additional arguments (detailed in the sections below) as slurm_apply.

Adding auxiliary data and functions

The function passed to slurm_apply can only receive atomic parameters stored within a data frame. Suppose we want instead to apply a function func to a list of complex R objects, obj_list. To use slurm_apply in this case, we can wrap func in an inline function that takes an integer parameter.

sjob <- slurm_apply(function(i) func(obj_list[[i]]), 
                    data.frame(i = seq_along(obj_list)),
                    add_objects = c("func", "obj_list"),
                    nodes = 2, cpus_per_node = 2)

The add_objects argument specifies the names of any R objects (besides the parameters data frame) that must be accessed by the function passed to slurm_apply. These objects are saved to a .RData file that is loaded on each cluster node prior to evaluating the function in parallel.

By default, all R packages attached to the current R session will also be attached (with library) on each cluster node, though this can be modified with the optional pkgs argument.

Configuring SLURM options

The slurm_options argument allows you to set any of the command line options (view list) recognized by the SLURM sbatch command. It should be formatted as a named list, using the long names of each option (e.g. "time" rather than "t"). Flags, i.e. command line options that are toggled rather than set to a particular value, should be set to TRUE in slurm_options. For example, the following code:

sjob <- slurm_apply(test_func, pars, 
                    slurm_options = list(time = "1:00:00", share = TRUE))

sets the command line options --time=1:00:00 --share.

Generating scripts for later submission

When working from a R session without direct access to the cluster, you can set submit = FALSE within slurm_apply. The function will create the _rslurm_[jobname] folder and generate the scripts and .RData files, without submitting the job. You may then copy those files to the cluster and submit the job manually by calling sbatch submit.sh from the command line.

How it works / advanced customization

As mentioned above, the slurm_apply function creates a job-specific folder. This folder contains the parameters data frame and (if applicable) the objects specified as add_objects, both saved in .RData files. The function also generates a R script (slurm_run.R) to be run on each cluster node, as well as a Bash script (submit.sh) to submit the job to SLURM.

More specifically, the Bash script creates a SLURM job array, with each cluster node receiving a different value of the SLURM_ARRAY_TASK_ID environment variable. This variable is read by slurm_run.R, which allows each instance of the script to operate on a different parameter subset and write its output to a different results file. The R script calls parallel::mcMap to parallelize calculations on each node.

Both slurm_run.R and submit.sh are generated from templates, using the whisker package; these templates can be found in the rslurm/templates subfolder in your R package library. There are two templates for each script, one for slurm_apply and the other (with the word single in its title) for slurm_call.

While you should avoid changing any existing lines in the template scripts, you may want to add #SBATCH lines to the submit.sh templates in order to permanently set certain SLURM command line options and thus customize the package to your particular cluster setup.