R/set_jobs.R

In yangjl/huskeR: Genomic and Genetic analysis Pipeline on HPC

#' \code{Set up array job on farm}
#'
#' Farm is a computer cluster running slurm system.
#' Note, bigmem mem=8000 per cpu, hi med low mem=25000 per cpu, serial mem=1500 per cpu.
#'
#' @param shid Relative or absolute path and file name of your shell code, i.e. CL_test.sh.
#' @param shcode The commands inside your sh file.
#' @param arrayjobs A character specify the number of array you try to run, i.e. 1-100.
#' @param wd Working directory, default=NULL. It will use your current directory.
#' @param jobid The job name show up in your sq NAME column.
#' @param email Your email address that farm will email to once the job was done or failed.
#' @param runinfo  Parameters specify the array job partition information.
#' A vector of c(TRUE, "bigmemh", "1"): 1) run or not, 2) -p partition name, and 3) --cpus.
#'
#' @return return a shell file.
#'
#' @examples
#' for(i in 1:10){
#'     shid <- paste0("slurm-script/run_", i, ".sh")
#'     command <- paste0("bedtools getfasta -name -tab -fi roast.chrom.", i, ".msa.in")
#'     cat(command, file=shid, sep="\n", append=FALSE)
#' }
#' shcode <- paste("module load bismark/0.14.3", "sh slurm-script/run_$SLURM_ARRAY_TASK_ID.sh", sep="\n")
#'
#' set_array_job(shid="slurm-script/run.sh", shcode=shcode,
#'               arrayjobs="1-10", wd=NULL, jobid="myjob", email=NULL,
#'               run = c(TRUE, "bigmemh", "8196", "1"))
#'
#' @export
set_array_job <- function(
  shid="largedata/GenSel/CL_test.sh", shcode="sh largedata/myscript.sh",
  arrayjobs="1-700", wd=NULL, jobid="myjob", email=NULL,
  runinfo=c(TRUE, "bigmemh", "1", "10G", "8:00:00")  ){

    #message(sprintf("###>>> cp from Introgression, tailored for pvpDiallel"))
    ##### setup working directory
    if(is.null(wd)){
       wd <- getwd()
    }
    dir.create("slurm-log", showWarnings = FALSE)
    sbath <- paste0(wd, "/slurm-log/")
    sbatho <- paste0(sbath, "testout-%j.txt")
    sbathe <- paste0(sbath, "err-%j.txt")

    #### parameters pass to slurm script
    cat(paste("#!/bin/bash -l"),
        #-D sets your project directory.
        #-o sets where standard output (of your batch script) goes.
        #-e sets where standard error (of your batch script) goes.
        #-J sets the job name.
        paste("#SBATCH -D", wd, sep=" "),
        paste("#SBATCH -o", sbatho, sep=" "),
        paste("#SBATCH -e", sbathe, sep=" "),
        paste("#SBATCH -J", jobid, sep=" "),
        paste0("#SBATCH --array=", arrayjobs),
        paste0("#SBATCH --mail-user=", email),
        paste("#SBATCH --mail-type=END"),
        paste("#SBATCH --mail-type=FAIL #email if fails"),

        "set -e",
        "set -u",
        "",
        #"module load gmap/2014-05-15",
        file=shid, sep="\n", append=FALSE);

    #### the sbatch code
    runinfo <- get_runinfo(runinfo)
    runcode <- paste0("sbatch -p ", runinfo[2], " --licenses=common --ntasks=", runinfo[3],
                      " --mem ", runinfo[4], " --time=", runinfo[5],
                      " ", shid)

    #### attach some sh scripts
    cat(shcode, file=shid, sep="\n", append=TRUE)
    if(runinfo[1]){
      message(runcode)
      system(runcode)
    }else{
      message(paste("###>>> In this path: cd ", wd, sep=""), "\n",
              paste("###>>> RUN:", runcode))
    }
}

#' @rdname set_array_job
#' @export
get_runinfo <- function(runinfo){
  ### determine memory based on partition
  run <- runinfo
  if(length(grep("med|hi|low", run[2])) > 0){
    mem <- 2600*as.numeric(run[3])
    runinfo <- c(run, mem)
  }else if(length(grep("bigmem", run[2])) > 0){
    mem <- 8196*as.numeric(run[3])
    runinfo <- c(run, mem)
  }else if(length(grep("serial", run[2])) > 0){
    mem <- 1500*as.numeric(run[3])
    runinfo <- c(run, mem)
  }
  return(runinfo)
}

#' \code{Set up one farm job}
#'
#' Farm is a computer cluster running slurm system.
#' Note, bigmem mem=8000/cpu, hi/med/low mem=25000/cpu, serial mem=1500/cpu.
#'
#' @param slurmsh Relative or absolute path and file name of your shell code, i.e. largedata/GenSel/CL_test.sh.
#' @param shcode The commands inside your sh file.
#' @param wd Working directory, default=NULL. It will use your current directory.
#' @param jobid The job name show up in your sq NAME column.
#' @param email Your email address that farm will email to once the job was done/failed.
#' @param runinfo  Parameters control the array job partition. c(TRUE, "bigmemh", "1", "2G", "8:00:00")
#' A vector of c(TRUE, "bigmemh", "1"): 1) run or not, 2) -p partition name, 3)  --ntasks.
#'
#'
#' @return a shell file.
#'
#' @examples
#' cmd <- paste0("snpconvert -a largedata/teo_updated/teo_raw_biallelic.hmp.txt",
#' " -i largedata/teo_updated/TeoCurated_20160308_AGPv2_flt_maf005m2.txt -s 12",
#' " -o largedata/teo_updated/teo_raw_biallelic_recoded_20160303_AGPv2.txt")
#'
#' set_farm_job(slurmsh = "largedata/scripts/run_snpconvert.sh",
#' shcode = cmd, wd = NULL, jobid = "snpconvert", email=NULL, runinfo=c(TRUE, "bigmemh", "1"))
#'
#' @export
set_slurm_job <- function(slurmsh="largedata/GenSel/CL_test.sh",
                         shcode="sh largedata/myscript.sh",
                         wd=NULL, jobid="myjob", email=NULL,
                         runinfo=c(TRUE, "bigmemh", "1", "2G", "8:00:00")){
  ##### setup working directory
  if(is.null(wd)){
    wd <- getwd()
  }
  dir.create("slurm-log", showWarnings = FALSE)
  sbath <- paste0(wd, "/slurm-log/")
  sbatho <- paste0(sbath, "testout-%j.txt")
  sbathe <- paste0(sbath, "err-%j.txt")
  sbathJ <- jobid

  #### parameters pass to slurm script
  cat(paste("#!/bin/bash"),
      #-D sets your project directory.
      #-o sets where standard output (of your batch script) goes.
      #-e sets where standard error (of your batch script) goes.
      #-J sets the job name.
      paste("#SBATCH -D", wd, sep=" "),
      paste("#SBATCH -o", sbatho, sep=" "),
      paste("#SBATCH -e", sbathe, sep=" "),
      paste("#SBATCH -J", sbathJ, sep=" "),
      paste0("#SBATCH --mail-user=", email),
      paste("#SBATCH --mail-type=END"),
      paste("#SBATCH --mail-type=FAIL #email if fails"),


      "set -e",
      "set -u",
      "",
      #"module load gmap/2014-05-15",
      file=slurmsh, sep="\n", append=FALSE);

  #### attach some sh scripts
  cat(shcode, file=slurmsh, sep="\n", append=TRUE)

  #### the sbatch code
  #runinfo <- get_runinfo(runinfo)
  runcode <- paste0("sbatch -p ", runinfo[2], " --licenses=common --ntasks=", runinfo[3],
                    " --mem=", runinfo[4],
                    " --time=", runinfo[5], " ", slurmsh)

  if(runinfo[1]){
    message(runcode)
    system(runcode)
  }else{
    message(paste("###>>> In this path: cd ", wd, sep=""), "\n",
            paste("###>>> RUN:", runcode))
  }
}