TORQUE 설치

0. 호스트 정보

A : gpucluster.kobic.kr

B : compute-0-0

A,B]# vi /etc/hosts

192.168.1.1     gpucluster.kobic.kr     gpucluster

192.168.1.2     compute-0-0.local       compute-0-0

1. B 호스트에 GPU Drivers, Toolkit, SDK 설치

http://ctrlcv.co.cc/entry/Install-NVIDIA-DRIVER-CUDA-Toolkit-SDK-%EC%84%A4%EC%B9%98CentOS-55

2..헤드노드에 torque 설치

다운로드 : http://clusterresources.com/downloads/torque

A # cd /usr/local/src

A # wget http://www.adaptivecomputing.com/resources/downloads/torque/torque-3.0.5.tar.gz

A # tar zxvfp torque-3.0.5.tar.gz

A # cd torque-3.0.5

### 컴파일 및 설치

A # ./configure --with-debug --enable-nvidia-gpus --enable-server --enable-clients --enable-docs --enable-mom --with-default-server=gpucluster.kobic.kr --with-scp && make && make install  

### 패키지 생성

A # make packages 

### 서버 기본 설정

A # ./torque.setup gpucluster.kobic.kr   

initializing TORQUE (admin: gpucluster.kobic.kr@gpucluster.kobic.kr)

Max open servers: 10239

Max open servers: 10239

### 노드 추가

A # vi /var/spool/torque/server_priv/nodes    

compute-0-0 np=16

### 데몬 실행

A # pbs_server

A # pbs_sched

### 데몬 재시작

A # killall pbs_server

A # pbs_server

A # killall pbs_sched

A # pbs_sched

### rc.local 추가

A # echo "/usr/local/sbin/pbs_server" >> /etc/rc.d/rc.local

A # echo "/usr/local/sbin/pbs_sched" >> /etc/rc.d/rc.local

### PBS 노드 확인

A # pbsnodes -a

A # qnodes

3. 계산 노드에 torque 설치

A # scp *.sh compute-0-0:/usr/local/src/

B # cd /usr/local/src

B # ./torque-package-mom-linux-x86_64.sh --install

B # ./torque-package-clients-linux-x86_64.sh --install

B # ./torque-package-doc-linux-x86_64.sh --install

B # ./torque-package-devel-linux-x86_64.sh --install

B # echo "/usr/local/sbin/pbs_mom" >> /etc/rc.d/rc.local

B # echo "\$pbsserver gpucluster.kobic.kr" > /var/spool/torque/mom_priv/config

B # echo "\$logevent 255" >> /var/spool/torque/mom_priv/config 

B # cat /var/spool/torque/server_name

gpucluster.kobic.kr

B # pbs_mom