待学习内容

1. 1.关于slurm系统
2. 2. slurm系统的sq命令
3. 3. array 任务 限制同时运行的任务总数
4. 4. 给每个文件分配一个文件夹
5. 5. from Dr. lan 跳过任务数量限制

1.关于slurm系统

1. 指定节点提任务 ：#SBATCH --nodelist=c1405
2. 杀残余进程 killall -9 vasp_std killall -9 vasp_gam

2. slurm系统的sq命令

#!/bin/bash
# A simple wrapper for squeue, requires bash 4.0+
# by Junhao Li@TCB

USAGE()
{
echo
echo "available arguments: "
echo "   -h | --help             print the help information"
echo "   -P | --show_partition   number (optional, default is 10)"
echo "   -j | --job_length       number (optional, default is 20)"
echo "   -p | --workidng_dir     number (optional, default is a full path)"
echo "   -a | --show_account     number (optional, default is 15)"
echo "   -A | --tcb_users"
echo "    *                      other arguments pass to squeue"
echo
}

resize >& /dev/null

[[ -z $1 ]] && squeue -u $USER -o ' %10i %15a %20j %8u %3t %10M %10l %4D%5C %20V %Z' && exit 0

squeueArgs=()
TCBUserList="lilu,user2,user3,user4" # replace the username here

while [[ $# -gt 0 ]]; do
    case $1 in 
    -P | --show_partition)
    showPartition="%10P"
    [[ "$2" -gt 0 ]] && showPartition="%${2}P" && shift 2 || shift 1
    ;;
    -U | --User_list) #user1,user2,.. or user1
    if [[ -z $showTCBUsers ]]; then
        if [[ "$2" -gt 0 ]]; then
            userList="-u ${2}"
            shift 2
        else
            echo "Warning: please give user after -u"
            shift 1
        fi
    else
        echo "-u could not be used with -A!"
        exit 1
    fi    
    ;;
    -j | --job_length)
    [[ "$2" -gt 0 ]] && jobLength="%${2}j" || echo "Warning: please give number after -j"
    shift 2
    ;;
    -p | -wd | --working_dir)
    showWorkingDir="%Z"
    [[ "$2" -gt 0 ]] && showWorkingDir="%${2}Z" && shift 2 || shift 1
    ;;
    -a | --show_account)
    showAccount="%15a"
    [[ "$2" -gt 0 ]] && showAccount="%${2}a" && shift 2 || shift 1
    ;;
    -A | --tcb_users)
    if [[ ! -z userList ]]; then
        showTCBUsers="-u $TCBUserList" 
    shift 1
    else
        echo "-A could not be used with -u!"
        exit 1
    fi
    ;;
    -s | --star | --start)
    showStartTime="--start"
    shift 1
    ;;
    -aAp | -apA | -Aap | -Apa | -paA | -pAa)
    showAccount="%15a"
    showTCBUsers="-u $TCBUserList"
    showWorkingDir="%Z"
    shift 1
    ;;
    -h | --help)
    USAGE
    exit 0
    ;;
    *) # Other argurement pass to squeue!
    squeueArgs=(${squeueArgs[*]} $1)
    shift 1
    ;;
    esac
done

[[ -z $jobLength ]] && jobLength="%20j"

#if [[ -z $userList ]] && [[ -z $showTCBUsers ]]; then
#    userToShow="-u $USER"
#elif [[ ! -z $userList ]]; then
#    userToShow=$userList
#elif [[ ! -z $showTCBUsers ]]; then
#    userToShow=$showTCBUsers
#fi

format=$(echo " %10i $showPartition $showAccount $jobLength %8u %3t %10M %10l %4D%5C %20V $showWorkingDir")

squeue $userToShow -o "$format" $showStartTime ${squeueArgs[@]}

3. array 任务 限制同时运行的任务总数

#SBATCH --array=1-100%10

1-100 指定了任务数组的范围，即从 1 到 100 的任务。

%10 限制了同时运行的任务数为 10

–array的数字范围只能在0-1001 ，1001是最大值

–array是重复提交固定次数，运行标头下面的内容，然后可以做一些限制

#!/bin/sh
#SBATCH  --job-name=scf
#SBATCH  --output=log.out.%A_%a
#SBATCH  --error=log.err.%A_%a
#SBATCH  --partition=intel6240r_192
#SBATCH  --nodes=1
#SBATCH  --ntasks=48
#SBATCH  --ntasks-per-node=48
#SBATCH  --array=0-1000%15
#SBATCH  --cpus-per-task=1
#SBATCH  --exclusive


dir=$((SLURM_ARRAY_TASK_ID + 1000))
cd $dir || { echo "Failed to enter directory $dir"; exit 1; }

source /work/env/intel2018

srun hostname | sort | uniq >> /tmp/nodefile.$$
NP=`srun hostname | wc -l`

mpirun -genv I_MPI_FABRICS shm:tcp -machinefile /tmp/nodefile.$$ -n $NP /work/software/vasp.6.1.0/vasp_std  > vasp.log 2>&1
#mpirun -genv I_MPI_FABRICS shm:ofa -machinefile /tmp/nodefile.$$ -n $NP /work/software/vasp.6.1.0/vasp_std  > vasp.log 2>&1  

rm -rf /tmp/nodefile.$$

4. 给每个文件分配一个文件夹

#!/bin/bash

# 初始化计数器
counter=0

# 获取当前目录下的所有文件
for file in *; do
    # 检查是否是文件（排除目录）
    if [ -f "$file" ]; then
        # 使用计数器作为文件夹名
        foldername="$counter"
        
        # 创建文件夹，以计数器命名
        mkdir -p "$foldername"
        
        # 移动文件到对应的文件夹
        mv "$file" "$foldername/"
        
        # 递增计数器
        counter=$((counter + 1))
    fi
done

5. from Dr. lan 跳过任务数量限制

VASPpath='/data/home/xieyu/workplace/songcn/opt/vasp.6.3.2_OPTCELL/bin'

outfile=flag_begin
workdir=/data/home/xieyu/workplace/lizhao/workplace/5-scf/pos

for i in ./*/
do
    if ls ${i}/${outfile} &> /dev/null; then
        echo "${i} calculated by others"
    else
        cd $i
        touch ${outfile}
        echo -n ${i}
        mpirun -n 56 ${VASPpath}/vasp_gam > vasp.log 2>&1
        echo '  done.'
        cd ${workdir}
    fi
done

---

转载请注明来源 有问题可通过github提交issue