OPENHPC搭建集群

目前搭建HPC的开源工具主要有RocksCluster, Xcat，Open-HPC。其中OpenHPC 是Linux基金开源的超算项目，致力于为高性能计算构建一个开源框架，适应学术研究的需求，为 HPC 环境创建一个开源框架，降低成本。

本次采用虚拟机实验，没有Infiniband,故先不介绍Infiniband的配置

主节点系统（CentOS-7.7）配置

1. 配置参数说明

环境变量	变量名称说明	模版参数
${sms_name}	主机名	server
${sms_ip}	主节点内网地址	192.168.130.100
$sms_eth_internal	主节点内网网卡名称	ens33
${eth_provision}	主节点与计算节点通信网卡名称	ens33
${internal_netmask}	内网子网掩码	255.255.255.0
${ntp_server}	本地时间同步服务器	192.168.130.100
${c_ip[0]}, {c_ip[1]} ,…	计算节点名称	192.168.130.[1-5]
${c_name[0]}, ${c_name[1]} ,…	计算节点名称	192.168.130.[1-5]
${sms_ipoib}	主节点Inifiniband网络地址	192.168.120.100
${ipoib_netmask}	主节点Infiniband网络子网掩码	255.255.255.0
${c_ipoib[0]},&{c_ipoib[1]},…	计算节点Infiniband网络IP地址	192.168.120.[1-5]
$compute_regex	计算节点名称正则表达式	node[1-5]
${compute_prefix}	计算节点名称前缀	node
${kargs}	Linux内核启动参数	net.ifnames=1

2. 添加hosts文件信任

	[root@server ~]#echo "192.168.130.1  node1" >> /etc/hosts
	[root@server ~]#echo "192.168.130.2  node2" >> /etc/hosts
	[root@server ~]#echo "192.168.130.3  node3" >> /etc/hosts
	[root@server ~]#echo "192.168.130.4  node4" >> /etc/hosts
	[root@server ~]#echo "192.168.130.5  node5" >> /etc/hosts

3. 关闭防火墙

	[root@server ~]#systemctl stop firewaled.sevice
	[root@server ~]#systemctl disabled firewaled.sevice

4. 关闭SELinux ；

	[root@server ~]#vi /etc/selinux/deisabled

5. 配置NTP服务 ；

	[root@server ~]#echo "server 192.168.130.100" >> /etc/ntp.conf
	[root@server ~]#systemctl enable ntpd.service

6. 配置内网网卡 ；

	[root@server ~]# vi /etc/sysconfig/network-scripts/ifcfg-ens33
	TYPE=Ethernet
	PROXY_METHOD=none
	BROWSER_ONLY=no
	BOOTPROTO=static
	DEFROUTE=yes
	IPV4_FAILURE_FATAL=no
	IPV6INIT=yes
	IPV6_AUTOCONF=yes
	IPV6_DEFROUTE=yes
	IPV6_FAILURE_FATAL=no
	IPV6_ADDR_GEN_MODE=stable-privacy
	NAME=ens33
	UUID=******************************
	DEVICE=ens33
	ONBOOT=yes
	IPADDR=192.168.130.100
	PREFIX=24
	IPV6_PRIVACY=no

7. ohpc-base,warewulf,pbs ；

安装open-HPC yum源 ohpc-release；

	[root@server ~]#yum install http://build.openhpc.community/OpenHPC:/1.3/CentOS_7/x86_64/ohpc-release-1.3-1.el7.x86_64.rpm
	[root@server ~]#yum -y install ohpc-base
	[root@server ~]#yum -y install ohpc-warewulf
	[root@server ~]#yum -y install pbspro-server-ohpc

管理节warewulf点配置

将/etc/warewulf/provision.conf文件中的 network device = eth1 改成内网网卡ens33

	[root@server ~]# vi /etc/warewulf/provision.conf
		# What is the default network device that the master will use to
		# communicate with the nodes?
		network device = ens33
		# Which DHCP server implementation should be used?
		dhcp server = isc
		
		# What is the TFTP root directory that should be used to store the
		# network boot images? By default Warewulf will try and find the
		# proper directory. Just add this if it can't locate it.
		#tftpdir = /var/lib/tftpboot
		
		# Automatically generate and manage a dynamnic_host virtual file
		# object in the datastore? This is useful for provisioning this
		# out to nodes so they always have a current /etc/hosts file.
		generate dynamic_hosts = yes
		
		# Should we manage and overwrite the local hostfile file on this
		# system? This will cause all node entries to be added
		# automatically to /etc/hosts.
		update hostfile = yes
		
		# If no cluster/domain is set on a node, should we add 'localdomain'
		# as the default domain
		use localdomain = yes
		
		# The default kernel arguments to pass to the nodes boot kernel
		default kargs = "net.ifnames=0 biosdevname=0 quiet"

编辑/etc/xinetd.d/tftp文件, 将disable = yes改为 disable = no

	[root@server ~]# vi /etc/xinetd.d/tftp
	# default: off
	# description: The tftp server serves files using the trivial file transfer \
	#       protocol.  The tftp protocol is often used to boot diskless \
	#       workstations, download configuration files to network-aware printers, \
	#       and to start the installation process for some operating systems.
	service tftp
	{
   
	        socket_type