Centos7安装pbs

机器配置

安装

配置好主机名和SSH互相免密

安装依赖

yum -y install libX11-devel libXt-devel libedit-devel libical-devel ncurses-devel postgresql-devel python-devel python-devel tcl-devel tk-devel swig expat-devel openssl-devel libXext libXft gcc autoconf automake libtool hwloc-devel rpm-build

pbs github地址

Releases · openpbs/openpbs (github.com)

wget https://github.com/openpbs/openpbs/releases/download/v19.1.3/pbspro_19.1.3.centos_7.zip
tar zxvf pbspro_19.1.3.centos_7.zip

设置环境变量 追加到/etc/profile

export PBS_SERVER=master
export PBS_EXEC=/opt/pbs
export PBS_HOME=/var/spool/pbs

master节点 安装 pbspro-server-19.1.3-0.x86_64.rpm

node节点 安装 pbspro-execution-19.1.3-0.x86_64.rpm

master节点和node节点依次启动pbs

systemctl start pbs

如果想让master节点也参与计算，将/etc/pbs.conf 中 PBS_START_MOM的值由0修改为1 ，其他节点psb.conf不用改变

配置

master节点其他配置

qmgr -c "set server flatuid=true"
qmgr -c "set server scheduling=true"
qmgr -c "s s job_history_enable=1"

master节点添加节点node01

qmgr -c "create node node01"

查看节点状态

pbsnodes -aSj

master节点创建队列big

qmgr -c "create queue big"

将node7添加到队列big

qmgr -c "set node node7 queue=big"

设置big队列资源限制

qmgr -c "set queue big resources_max.ncpus = 80"
qmgr -c "set queue big resources_max.mem = 503"
qmgr -c "set queue big resources_max.nodect = 1"

启用big队列

qmgr -c "set queue big enabled = true"
qmgr -c "set queue big started = true"

查看big队列参数

qstat -Q -f big

注意 普通用户提交作业前需要配置和计算节点间的SSH互相免密

配置gpu队列

管理节点 /var/spool/pbs/sched_priv/sched_config 新增

resources: "ncpus, mem, arch, host, vnode, aoe, eoe, ngpus, gpu_id"

然后重启pbs

service pbs restart 

设置下gpu节点的资源

# qmgr
set node node1 resources_available.ncpus=48, resources_available.mem=100gb, resources_available.ngpus=2 

创建gpu队列qgpu 并重启

# qmgr
create queue qgpu
set queue qgpu queue_type = Execution 
set queue qgpu resources_default.nodect = 1
set queue qgpu resources_default.nodes = node1
set queue qgpu resources_default.ngpus = 2
set queue qgpu resources_default.ncpus = 48
set queue qgpu enabled = True 
set queue qgpu started = True
exit
# service pbs restart