BeeOND部署文档
- 1、安装beeOND
- 2、配置安装目录及ssh免密
- 3、停止beeOND
- 4、其他工具使用
- 5、性能测试
beeOND是beeGFS为基础一种快速组建beegfs机器的一个工具,主要用于GPU节点的ssd硬盘,组建成一个高速的beegfs存储池(和原beegfs存储挂载相互不受影响)
1、安装beeOND
所有的节点配置yum源,在线安装,,会同时安装beegfs
[root@beeond1 ~]# cat /etc/yum.repos.d/beegfs.repo
[beegfs]
name=BeeGFS 7 (RHEL7)
If you have an active BeeGFS support contract, use the alternative URL below
to retrieve early updates. Replace username/password with your account for
the BeeGFS customer login area.
baseurl=https://www.beegfs.io/release/beegfs_7_1/dists/rhel7
gpgkey=https://www.beegfs.io/release/beegfs_7_1/gpg/RPM-GPG-KEY-beegfs
gpgcheck=0
enabled=1
[root@beeond1 ~]# yum install beeond -y
2、配置安装目录及ssh免密
- nodefile:包含所有运行BeeOND实例主机名的文件,所有在这个文件上的主机都会成为这个新的文件系统实例的client以及server
- /data/beeond:BeeOND实例在每个节点上存储原始数据的路径(每台机都要)
- /mnt/beeond:BeeOND实例在每个节点的挂载点(每台机都要)
将所有的节点配置ssh免密
nodefile只需要在其中一个节点上(注意nodefile只能使用ip或者主机名,不能同时使用,否则会有问题,默认meta、mgmtd都会部署在nodefile里面的第一台机)
使用命令启动一个beeond实例,(-m 2表示部署两个meta节点)
beeond start -n nodefile -d /data/beeond -c /mnt/beeond -m 2
[root@beeond1 ~]# beeond start -n nodefile -d /data/beeond -c /mnt/beeond -m 2
INFO: Using status information file: /var/tmp/beeond.tmp
INFO: Checking reachability of host beeond1
INFO: Checking reachability of host beeond2
INFO: Checking reachability of host beegfs-mgmtd
INFO: Number of storage servers automatically set to 3
INFO: Starting beegfs-mgmtd processes
INFO: Management daemon log: /var/log/beegfs-mgmtd_20220224-110319.log
INFO: Starting beegfs-mgmtd on host: beeond1
INFO: Starting beegfs-storage processes
INFO: Storage server log: /var/log/beegfs-storage_20220224-110319.log
INFO: Starting beegfs-storage on host: beeond1
INFO: Starting beegfs-storage on host: beeond2
INFO: Starting beegfs-storage on host: beegfs-mgmtd
INFO: Starting beegfs-meta processes
INFO: Metadata server log: /var/log/beegfs-meta_20220224-110319.log
INFO: Starting beegfs-meta on host: beeond1
INFO: Starting beegfs-meta on host: beeond2
INFO: Starting beegfs-client processes
INFO: Client log: /var/log/beegfs-client_20220224-110319.log
INFO: Starting beegfs-client on host: beeond1
INFO: Starting beegfs-client on host: beeond2
INFO: Starting beegfs-client on host: beegfs-mgmtd
[root@beeond1 ~]# df -h
Filesystem Size Used Avail Use% Mounted on
devtmpfs 899M 0 899M 0% /dev
tmpfs 910M 0 910M 0% /dev/shm
tmpfs 910M 9.7M 901M 2% /run
tmpfs 910M 0 910M 0% /sys/fs/cgroup
/dev/mapper/centos-root 17G 1.7G 16G 10% /
/dev/sdb 20G 303M 19G 2% /data/beeond
/dev/sda1 1014M 190M 825M 19% /boot
tmpfs 182M 0 182M 0% /run/user/0
beegfs_ondemand 57G 2.4G 54G 5% /mnt/beeond
3、停止beeOND
beeond stop -n nodefile (后面可以不用加参数,如果后面加了-d的参数,会把所有的数据清除掉)
[root@beeond1 ~]# beeond stop -n nodefile
INFO: Using status information file: /var/tmp/beeond.tmp
INFO: Checking reachability of host beeond1
INFO: Checking reachability of host beeond2
INFO: Checking reachability of host beegfs-mgmtd
INFO: Unmounting file system on host: beeond1
/opt/beegfs/sbin/…/lib/beegfs-ondemand-stoplocal: line 83: fuser: command not found
INFO: Unmounting file system on host: beeond2
/opt/beegfs/sbin/…/lib/beegfs-ondemand-stoplocal: line 83: fuser: command not found
INFO: Unmounting file system on host: beegfs-mgmtd
/opt/beegfs/sbin/…/lib/beegfs-ondemand-stoplocal: line 83: fuser: command not found
INFO: Stopping remaining processes on host: beeond1
INFO: Stopping remaining processes on host: beeond2
INFO: Stopping remaining processes on host: beegfs-mgmtd
INFO: Deleting status file on host: beeond1
INFO: Deleting status file on host: beeond2
INFO: Deleting status file on host: beegfs-mgmtd
4、其他工具使用
使用beegfs-ctl查看BeeOND实例信息
[root@beeond1 ~]# beegfs-ctl --mount=/mnt/beeond --listnodes --nodetype=storage
beeond1 [ID: 1]
beeond2 [ID: 2]
beegfs-mgmtd [ID: 3]
[root@beeond1 ~]# beegfs-ctl --mount=/mnt/beeond --listnodes --nodetype=meta
beeond1 [ID: 1]
beeond2 [ID: 2]
beegfs-mgmtd [ID: 3]
[root@beeond1 ~]# beegfs-ctl --mount=/mnt/beeond --listnodes --nodetype=mgmt
beeond1 [ID: 1]
beegfs-cp 将数据拷入或者拷出BeeOND
beegfs-cp copy -n nodefile /projects/dataset01 /projects/dataset02 /mnt/beeond
参数说明:
nodefile:包含所有主机名
/projects/dataset01 & /projects/dataset02:要拷贝数据的源路径(支持一个或者多个源文件或者源目录)
/mnt/beeond:目标路径
5、性能测试
一般使用fio测试,分别测试4K和1M的块大小,测试iops和带宽
性能和storage节点数有很大的关系,其中部署多个meta节点可以避免meta的单点问题,但是在测试过程中mgmtd只有一个节点,无法部署多节点,当mgmtd宕机时,会出现整个beeOND集群无法使用,存在多个meta或者storage时,宕机不会影响集群,但是同时存在一个致命的bug,当某一节点宕掉后,不会自动加回集群,需要重启整个集群才行,并不适用于生产环境。