꼬반 Blog

[오픈소스] GOQ 구축 (goq: a queuing and job management system fir for the cloud. Written in GO (golang)).

설치 및 구축 프로세스

마스터 노드, 워커 노드 셋팅

VM에 마스터 노드, 워커 노드 2대 생성 후 CentOS 7.2 설치

설치 후 각 서버는 selinux 설정을 permissive 로 변경

워커 노드는 firewalld.service 를 stop 및 disable 시킴

마스터 노드에 eth0, eth1 connection 설정

각 서버 네임서버 셋팅 (210.218.222.100)

dhcpd 설치 및 셋팅 (yum 으로 설치 후 아래처럼 dhcpd.conf 수정)

option space PXE;
option PXE.mtftp-ip    code 1 = ip-address;
option PXE.mtftp-cport code 2 = unsigned integer 16;
option PXE.mtftp-sport code 3 = unsigned integer 16;
option PXE.mtftp-tmout code 4 = unsigned integer 8;
option PXE.mtftp-delay code 5 = unsigned integer 8;
option arch code 93 = unsigned integer 16; # RFC4578

subnet 10.10.10.0 netmask 255.255.255.0 {
    option routers 10.10.10.10;
    range dynamic-bootp 10.10.10.200 10.10.10.250;
    allow booting;
    allow bootp;
    next-server 10.10.10.10;
    filename "pxelinux.0";
}
:wq!
# systemctl enable dhcpd
# systemctl start dhcpd

방화벽 설정

1. 마스커레이드 설정

2. 마스터서버와 워커노드가 통신을 하기위해 해당 대역에 대한 룰 설정

# firewall-cmd --permanent --add-masquerade
# firewall-cmd --permanent --add-rich-rule='rule family="ipv4" source address="10.10.10.0/24" accept'
# firewall-cmd --reload
  • 워커 노드에 접속해서 네트워크 설정 상태 확인 (마스터 및 다른 노드와 통신이 되는지 확인, nslookup으로 외부 인터넷과 연결이 되는지 확인)
  • 각 서버 OS 최신 패치 yum update

GO 설치

  • Go는 https://golang.org/dl/ 에서 작성일 기준 1.6.2 버전 다운 (동작 확인)
  • /share/apps/ 폴더에 다운 및 압축 해제 (미리 빌드된 버전 사용)
  • 사용자 환경설정에 GOROOT, GOPATH 설정 (root 기준)
vim ~/.bashrc
export GOROOT=/share/apps/go
export PATH=$PATH:$GOROOT/bin
export GOPATH=/share/workspace
export PATH=$PATH:$GOPATH/bin
export GOQ_HOME=/root
추가 후
:wq!
# go version
go version go1.6.2 linux/amd64
  • GOPATH는 워크스페이스 개념
  • GOQ_HOME은 goq init 시 환경 파일이 저장될 위치

goq 설치

github src : https://github.com/glycerine/goq

# go get -t -u github.com/glycerine/mangos/compat
# go get -u -t github.com/glycerine/goq
# cd $GOPATH/src/github.com/glycerine/goq; make; go test -v
  • 3번째 단계에서 goq/vendor/github.com 안에 보면 go-mangos 폴더이름을 gdamore 로 변경
  • 이유는 goq 내 소스에서 해당 mangos 패키지를 참조하는데 github.com/gdamore/mangos 로 되어 있음. (아마도 기존 경로로 보임, 현재 mangos 는 github.com/go-mangos/mangos 로 변경되어 있음)
  • 소스 내 참조를 모두 변경하기 어려우니 해당 폴더명만 변경 후 make 하면 에러가 나지 않고 build 가 완료 됨.
  • 이후 각 서버 별로 테스트 확인

마스터, 워커 설정

  • 먼저 마스터에서 설정
# goq init
  • 이후 GOQ_HOME 에 .goq 폴더 생성 됨. 이 안에 key와 환경파일이 생성.
# cd $GOQ_HOME
# vim serverloc
  • 서버 IP 셋팅 (디폴트는 eth0 ip 를 사용하므로 마스터, 워커가 사용하는 대역으로 변경)
export GOQ_JSERV_IP=10.10.10.10
:wq!
  • 다른 옵션은 디폴트 사용
  • 위 작업 완료 후 마스터의 .goc 폴더를 각 워커 GOQ_HOME에 복사
마스터
$ cd $GOQ_HOME
$ goq init     # only needed once.
$ nohup goq serve &   # start the central server

워커
$ ssh computenode
$ for i in $(seq 1 $(cat /proc/cpuinfo |grep processor|wc -l)); do 
  /usr/bin/nohup goq work forever & done

잡 서밋
$ cd somewhere/where/the/job/wants/to/start
# start by doing 'goq sub' on the same machine 
# that 'goq serve' was launched on. Just to learn the system.
$ goq sub ./myjobscript  

테스트
마스터에서 확인 
root      4060  2959  0 14:33 pts/0    00:00:00 goq serve

워커 1
root     12153 11924  0 14:36 pts/0    00:00:00 goq work forever
root     12154 11924  0 14:36 pts/0    00:00:00 goq work forever

워커 2
root     12215 12004  0 14:36 pts/0    00:00:00 goq work forever
root     12216 12004  0 14:36 pts/0    00:00:00 goq work forever

테스트 잡 서밋
[root@goq .goq]# goq sub /share/workspace/src/github.com/glycerine/goq/bin/sleep20.sh
[pid 5655] submitted job 5 to server at 'tcp://10.10.10.10:1776'.
[root@goq .goq]# goq stat
[pid 5661] stats for job server 'tcp://10.10.10.10:1776':
runQlen=1
waitingJobs=0
waitingWorkers=3
jservPid=4060
finishedJobsCount=3
droppedBadSigCount=0
cancelledJobCount=0
nextJobId=6
jservIP=10.10.10.10
jservPort=1776
badNonceCount=0
maxShow=10
runq 000000   runtime: < 1 heartbeat RunningJob[jid 5] = '/share/workspace/src/github.com/glycerine/goq/bin/sleep20.sh '   on worker 'tcp://10.10.10.201:39504'/pid:0. Lastping: none.   
finished: [jid 1] total-time: 20.006763604s. done: 2016-06-21 14:14:45.162962843 +0900 KST. cmd: '/share/workspace/src/github.com/glycerine/goq/bin/sleep20.sh []' finished on worker 'tcp://192.168.150.73:46641'/pid:3531.  . Err: ''
finished: [jid 3] total-time: 145.624801ms. done: 2016-06-21 14:37:12.110080928 +0900 KST. cmd: '/share/workspace/src/github.com/glycerine/goq/bin/good.sh []' finished on worker 'tcp://10.10.10.200:45736'/pid:12168.  . Err: ''
finished: [jid 4] total-time: 20.019386787s. done: 2016-06-21 14:38:33.327349501 +0900 KST. cmd: '/share/workspace/src/github.com/glycerine/goq/bin/sleep20.sh []' finished on worker 'tcp://10.10.10.200:43298'/pid:12172.  . Err: ''
--- goq security status---
summary-bad-signature-msgs: 0
summary-bad-nonce-msg: 0
--- goq progress status ---
summary-jobs-running: 1
summary-jobs-waiting: 0
summary-known-jobs: 1
summary-workers-waiting: 3
summary-cancelled-jobs: 0
summary-jobs-finished: 3
--- goq end status at time: 2016-06-21 15:44:40.540185547 +0900 KST ---
[root@goq .goq]# goq stat
[pid 5667] stats for job server 'tcp://10.10.10.10:1776':
runQlen=0
waitingJobs=0
waitingWorkers=4
jservPid=4060
finishedJobsCount=4
droppedBadSigCount=0
cancelledJobCount=0
nextJobId=6
jservIP=10.10.10.10
jservPort=1776
badNonceCount=0
maxShow=10
finished: [jid 1] total-time: 20.006763604s. done: 2016-06-21 14:14:45.162962843 +0900 KST. cmd: '/share/workspace/src/github.com/glycerine/goq/bin/sleep20.sh []' finished on worker 'tcp://192.168.150.73:46641'/pid:3531.  . Err: ''
finished: [jid 3] total-time: 145.624801ms. done: 2016-06-21 14:37:12.110080928 +0900 KST. cmd: '/share/workspace/src/github.com/glycerine/goq/bin/good.sh []' finished on worker 'tcp://10.10.10.200:45736'/pid:12168.  . Err: ''
finished: [jid 4] total-time: 20.019386787s. done: 2016-06-21 14:38:33.327349501 +0900 KST. cmd: '/share/workspace/src/github.com/glycerine/goq/bin/sleep20.sh []' finished on worker 'tcp://10.10.10.200:43298'/pid:12172.  . Err: ''
finished: [jid 5] total-time: 20.031627872s. done: 2016-06-21 15:44:58.280047843 +0900 KST. cmd: '/share/workspace/src/github.com/glycerine/goq/bin/sleep20.sh []' finished on worker 'tcp://10.10.10.201:39504'/pid:12970.  . Err: ''
--- goq security status---
summary-bad-signature-msgs: 0
summary-bad-nonce-msg: 0
--- goq progress status ---
summary-jobs-running: 0
summary-jobs-waiting: 0
summary-known-jobs: 0
summary-workers-waiting: 4
summary-cancelled-jobs: 0
summary-jobs-finished: 4
--- goq end status at time: 2016-06-21 15:45:00.384001311 +0900 KST ---
  • 해당 잡 스크립트
#!/bin/bash
echo "sleep.sh begins sleeping for 20 seconds"
sleep 20
echo "sleep.sh done"

워커 2번에서 실행이 된걸 확인할 수 있음.

결과
/root/.goq/o
out.00005 # jid

sleep.sh begins sleeping for 20 seconds
sleep.sh done
  • nfs 공유폴더 설정 및 hpl 벤치를 위한 설정
** 마스터 서버
# vim /etc/exports
/BiO 10.10.10.0/24(rw,sync,no_root_squash)

# systemctl start rpcbind
# systemctl start nfs-server
# systemctl start nfs-idmap
# systemctl start nfs-lock
# systemctl enable nfs-server

** 노드
# showmount -e 10.10.10.10
# mount -t nfs 10.10.10.10:/BiO /BiO

# vim /etc/fstab
10.10.10.10:/BiO /BiO defaults 0 0

** hpl 벤치 준비 (failed...)
** ATLAS 설치 및 install
** hpl 은 추후 다시 설치 시도...
  • 테스트
  • /BiO/test
  • 1 ~ 9 데이터셋
  • run.sh - 1번부터 9번을 읽어서 그 내용을 result 에 저장, 단 5초 대기 시간 있음
#!/bin/bash
echo "sleep 5"
sleep 5
cat $1 >> result
cat $1
  • job.sh - 단순한 반복문으로 실행. 각 파일마다 5초 딜레이 걸리므로 45초 후 작업이 완료
#!/bin/bash
for i in $(seq 1 9); do 
  ./run.sh $i
done
  • job2.sh - 해당 잡을 goq 로 서밋시킴. 이때 4개의 워커가 서밋된 잡을 실행. 즉 한번에 4개씩 처리. 15초에 작업 완료
#!/bin/bash
for i in $(seq 1 9); do 
  goq sub ./run.sh $i
done

[root@goq test]# goq stat
[pid 8221] stats for job server 'tcp://10.10.10.10:1776':
runQlen=0
waitingJobs=0
waitingWorkers=4
jservPid=6641
finishedJobsCount=22
droppedBadSigCount=0
cancelledJobCount=0
nextJobId=24
jservIP=10.10.10.10
jservPort=1776
badNonceCount=0
maxShow=10
finished: [jid 14] total-time: 5.006814286s. done: 2016-06-22 16:46:53.858905226 +0900 KST. cmd: './run.sh [9]' finished on worker 'tcp://10.10.10.200:38134'/pid:9547.  . Err: ''
finished: [jid 15] total-time: 5.009640992s. done: 2016-06-22 16:52:28.8312227 +0900 KST. cmd: './run.sh [1]' finished on worker 'tcp://10.10.10.200:52480'/pid:9678.  . Err: ''
finished: [jid 16] total-time: 5.009856439s. done: 2016-06-22 16:52:28.822657713 +0900 KST. cmd: './run.sh [2]' finished on worker 'tcp://10.10.10.201:54694'/pid:9931.  . Err: ''
finished: [jid 17] total-time: 5.014137871s. done: 2016-06-22 16:52:28.827363219 +0900 KST. cmd: './run.sh [3]' finished on worker 'tcp://10.10.10.201:54228'/pid:9933.  . Err: ''
finished: [jid 18] total-time: 5.005063449s. done: 2016-06-22 16:52:28.866128228 +0900 KST. cmd: './run.sh [4]' finished on worker 'tcp://10.10.10.200:38134'/pid:9681.  . Err: ''
finished: [jid 19] total-time: 5.007172282s. done: 2016-06-22 16:52:33.867240633 +0900 KST. cmd: './run.sh [5]' finished on worker 'tcp://10.10.10.200:52480'/pid:9686.  . Err: ''
finished: [jid 20] total-time: 5.014624321s. done: 2016-06-22 16:52:33.858707659 +0900 KST. cmd: './run.sh [6]' finished on worker 'tcp://10.10.10.201:54694'/pid:9941.  . Err: ''
finished: [jid 21] total-time: 5.015421181s. done: 2016-06-22 16:52:33.863655635 +0900 KST. cmd: './run.sh [7]' finished on worker 'tcp://10.10.10.201:54228'/pid:9942.  . Err: ''
finished: [jid 22] total-time: 5.011072741s. done: 2016-06-22 16:52:33.896002578 +0900 KST. cmd: './run.sh [8]' finished on worker 'tcp://10.10.10.200:38134'/pid:9691.  . Err: ''
finished: [jid 23] total-time: 5.002196052s. done: 2016-06-22 16:52:38.898022316 +0900 KST. cmd: './run.sh [9]' finished on worker 'tcp://10.10.10.200:52480'/pid:9696.  . Err: ''
--- goq security status---
summary-bad-signature-msgs: 0
summary-bad-nonce-msg: 0
--- goq progress status ---
summary-jobs-running: 0
summary-jobs-waiting: 0
summary-known-jobs: 0
summary-workers-waiting: 4
summary-cancelled-jobs: 0
summary-jobs-finished: 22
--- goq end status at time: 2016-06-22 17:43:52.258494225 +0900 KST ---
[root@goq test]#

이후 진행 방안

  • bwa 테스트 완료
  • 18개 fasta 파일 bwa 테스트 진행
run.sh (/BiO/bwa_data/)
#!/bin/bash
for i in *.fasta
do
    #echo $i
    goq sub /BiO/apps/bwa-0.7.15/bwa index $i
done

[root@goq bwa_data]# goq stat
[pid 5300] stats for job server 'tcp://10.10.10.10:1776':
runQlen=0
waitingJobs=0
waitingWorkers=4
jservPid=6641
finishedJobsCount=77
droppedBadSigCount=0
cancelledJobCount=0
nextJobId=79
jservIP=10.10.10.10
jservPort=1776
badNonceCount=0
maxShow=10
finished: [jid 70] total-time: 25.652276979s. done: 2016-06-24 11:26:47.437507654 +0900 KST. cmd: '/BiO/apps/bwa-0.7.15/bwa [index dmel-3R-chromosome-r5.37.fasta]' finished on worker 'tcp://10.10.10.201:54228'/pid:25409.  . Err: ''
finished: [jid 72] total-time: 2m57.995945527s. done: 2016-06-24 11:29:38.320234035 +0900 KST. cmd: '/BiO/apps/bwa-0.7.15/bwa [index dmel-all-chromosome-r5.37.fasta]' finished on worker 'tcp://10.10.10.200:52480'/pid:25095.  . Err: ''
finished: [jid 67] total-time: 5m20.065814901s. done: 2016-06-24 11:31:20.588482645 +0900 KST. cmd: '/BiO/apps/bwa-0.7.15/bwa [index dmel-2R-aligned-r5.37.fasta]' finished on worker 'tcp://10.10.10.201:54694'/pid:25399.  . Err: ''
finished: [jid 74] total-time: 2m32.706963928s. done: 2016-06-24 11:32:11.06134462 +0900 KST. cmd: '/BiO/apps/bwa-0.7.15/bwa [index dmel-all-gene_extended2000-r5.37.fasta]' finished on worker 'tcp://10.10.10.200:52480'/pid:25125.  . Err: ''
finished: [jid 76] total-time: 5.257377818s. done: 2016-06-24 11:32:16.329604209 +0900 KST. cmd: '/BiO/apps/bwa-0.7.15/bwa [index dmel-U-chromosome-r5.37.fasta]' finished on worker 'tcp://10.10.10.200:52480'/pid:25209.  . Err: ''
finished: [jid 69] total-time: 6m56.254709913s. done: 2016-06-24 11:33:07.201563569 +0900 KST. cmd: '/BiO/apps/bwa-0.7.15/bwa [index dmel-3R-aligned-r5.37.fasta]' finished on worker 'tcp://10.10.10.200:38134'/pid:25088.  . Err: ''
finished: [jid 78] total-time: 14.240585455s. done: 2016-06-24 11:33:21.461607565 +0900 KST. cmd: '/BiO/apps/bwa-0.7.15/bwa [index dmel-X-chromosome-r5.37.fasta]' finished on worker 'tcp://10.10.10.200:38134'/pid:25214.  . Err: ''
finished: [jid 77] total-time: 4m4.989334558s. done: 2016-06-24 11:36:21.328854204 +0900 KST. cmd: '/BiO/apps/bwa-0.7.15/bwa [index dmel-X-aligned-r5.37.fasta]' finished on worker 'tcp://10.10.10.200:52480'/pid:25211.  . Err: ''
finished: [jid 75] total-time: 6m35.538950491s. done: 2016-06-24 11:37:56.176649363 +0900 KST. cmd: '/BiO/apps/bwa-0.7.15/bwa [index dmel-all-predicted-r5.37.fasta]' finished on worker 'tcp://10.10.10.201:54694'/pid:25542.  . Err: ''
finished: [jid 73] total-time: 11m17.50777801s. done: 2016-06-24 11:38:04.953963217 +0900 KST. cmd: '/BiO/apps/bwa-0.7.15/bwa [index dmel-all-clones-r5.37.fasta]' finished on worker 'tcp://10.10.10.201:54228'/pid:25415.  . Err: ''
--- goq security status---
summary-bad-signature-msgs: 0
summary-bad-nonce-msg: 0
--- goq progress status ---
summary-jobs-running: 0
summary-jobs-waiting: 0
summary-known-jobs: 0
summary-workers-waiting: 4
summary-cancelled-jobs: 0
summary-jobs-finished: 77
--- goq end status at time: 2016-06-24 13:19:11.903386522 +0900 KST ---
[root@goq bwa_data]#
반응형

Article By 꼬반

*certificate* : VCP 5(2012), RHCSA 7 (2014), RHCE 7 (2015), RHCSA in REDHAT OpenStack(2017) *development language* : Javascript, NodeJS, Golang, html5, css3, shell script *middle ware* : NGINX, Apache, Tomcat, Docker, Docker Swarm, Mesos, Kubernetes, HCI,

Discuss about post