作者：徐凯
这篇文章主要介绍了如何用三台虚拟机搭建一套Ceph分布式系统，步骤简洁但不失准确性。环境清理一小节可以解决绝大多数部署不成功的问题，最后一节介绍了常用的Ceph操作，希望能给刚搭建环境的同学一些帮助。
环境
三台装有CentOS 7的主机，每台主机有三个磁盘(虚拟机磁盘要大于100G),详细信息如下：
[root@ceph-1~]# cat /etc/redhat-release
CentOS Linux release 7.2.1511 (Core)
[root@ceph-1 ~]# lsblk
NAME MAJ:MINRM SIZE RO TYPE MOUNTPOINT
sda 8:0 0 128G 0 disk
├─sda1 8:1 0 500M 0 part /boot
└─sda2 8:2 0 127.5G 0 part
├─centos-root 253:0 0 50G 0 lvm /
├─centos-swap 253:1 0 2G 0 lvm [SWAP]
└─centos-home 253:2 0 75.5G 0 lvm /home
sdb 8:16 0 2T 0 disk
sdc 8:32 0 2T 0 disk
sdd 8:48 0 2T 0 disk
sr0 11:0 1 1024M 0 rom
[root@ceph-1 ~]# cat /etc/hosts
..
192.168.57.222ceph-1
192.168.57.223ceph-2
192.168.57.224 ceph-3
集群配置如下：
| 主机 | IP | 功能 |
|---|---|---|
| ceph-1 | 192.168.57.222 | deploy、mon_1、osd_3 |
| ceph-2 | 192.168.57.223 | mon_1、 osd_3 |
| ceph-3 | 192.168.57.224 | mon_1 、osd_3 |
环境清理
如果之前部署失败了，不必删除ceph客户端，或者重新搭建虚拟机，只需要在每个节点上执行如下指令即可将环境清理至刚安装完ceph客户端时的状态！强烈建议在旧集群上搭建之前清理干净环境，否则会发生各种异常情况。
psaux|grep ceph |awk '{print $2}'|xargs kill -9
ps -ef|grep ceph
#确保此时所有ceph进程都已经关闭！！！如果没有关闭，多执行几次。
umount /var/lib/ceph/osd/*
rm -rf /var/lib/ceph/osd/*
rm -rf /var/lib/ceph/mon/*
rm -rf /var/lib/ceph/mds/*
rm -rf /var/lib/ceph/bootstrap-mds/*
rm -rf /var/lib/ceph/bootstrap-osd/*
rm -rf /var/lib/ceph/bootstrap-rgw/*
rm -rf /var/lib/ceph/tmp/*
rm -rf /etc/ceph/*
rm -rf /var/run/ceph/*
yum源及ceph的安装
需要在每个主机上执行以下指令
yumclean all
rm -rf /etc/yum.repos.d/*.repo
wget -O /etc/yum.repos.d/CentOS-Base.repo http://mirrors.aliyun.com/repo/Centos-7.repo
wget -O /etc/yum.repos.d/epel.repo http://mirrors.aliyun.com/repo/epel-7.repo
sed -i '/aliyuncs/d' /etc/yum.repos.d/CentOS-Base.repo
sed -i '/aliyuncs/d' /etc/yum.repos.d/epel.repo
sed -i 's/$releasever/7/g' /etc/yum.repos.d/CentOS-Base.repo
增加ceph的源:
vim/etc/yum.repos.d/ceph.repo
添加以下内容：
[ceph]
name=ceph
baseurl=http://mirrors.163.com/ceph/rpm-jewel/el7/x86_64/
gpgcheck=0
[ceph-noarch]
name=cephnoarch
baseurl=http://mirrors.163.com/ceph/rpm-jewel/el7/noarch/
gpgcheck=0
安装ceph客户端：
yummakecache
yum install ceph ceph-radosgw rdate -y
关闭selinux&firewalld
sed-i 's/SELINUX=.*/SELINUX=disabled/' /etc/selinux/config
setenforce 0
systemctl stop firewalld
systemctl disable firewalld
同步各个节点时间：
yum-y install rdate
rdate -s time-a.nist.gov
echo rdate -s time-a.nist.gov >> /etc/rc.d/rc.local
chmod +x /etc/rc.d/rc.local
开始部署
在部署节点(ceph-1)安装ceph-deploy，下文的部署节点统一指ceph-1：
[root@ceph-1~]# yum -y install ceph-deploy
[root@ceph-1 ~]# ceph-deploy --version
1.5.34
[root@ceph-1 ~]# ceph -v
ceph version 10.2.2 (45107e21c568dd033c2f0a3107dec8f0b0e58374)
在部署节点创建部署目录并开始部署：
[root@ceph-1~]# cd
[root@ceph-1 ~]# mkdir cluster
[root@ceph-1 ~]# cd cluster/
[root@ceph-1 cluster]# ceph-deploy new ceph-1 ceph-2 ceph-3
如果之前没有ssh-copy-id到各个节点，则需要输入一下密码，过程log如下：
[ceph_deploy.conf][DEBUG] found configuration file at: /root/.cephdeploy.conf
[ceph_deploy.cli][INFO ] Invoked (1.5.34): /usr/bin/ceph-deploy newceph-1 ceph-2 ceph-3
[ceph_deploy.cli][INFO ] ceph-deploy options:
[ceph_deploy.cli][INFO ] username : None
[ceph_deploy.cli][INFO ] func : <function new at 0x7f91781f96e0>
[ceph_deploy.cli][INFO ] verbose : False
[ceph_deploy.cli][INFO ] overwrite_conf : False
[ceph_deploy.cli][INFO ] quiet : False
[ceph_deploy.cli][INFO ] cd_conf : <ceph_deploy.conf.cephdeploy.Conf instance at 0x7f917755ca28>
[ceph_deploy.cli][INFO ] cluster : ceph
[ceph_deploy.cli][INFO ] ssh_copykey : True
[ceph_deploy.cli][INFO ] mon :['ceph-1', 'ceph-2', 'ceph-3']
..
..
ceph_deploy.new][WARNIN] could not connect via SSH
[ceph_deploy.new][INFO ] will connect again with password prompt
The authenticity of host 'ceph-2 (192.168.57.223)' can't be established.
ECDSA key fingerprint is ef:e2:3e:38:fa:47:f4:61:b7:4d:d3:24:de:d4:7a:54.
Are you sure you want to continue connecting (yes/no)? yes
Warning: Permanently added 'ceph-2,192.168.57.223' (ECDSA) to the list of knownhosts.
root
root@ceph-2's password:
[ceph-2][DEBUG ] connected to host: ceph-2
..
..
[ceph_deploy.new][DEBUG ] Resolving host ceph-3
[ceph_deploy.new][DEBUG ] Monitor ceph-3 at 192.168.57.224
[ceph_deploy.new][DEBUG ] Monitor initial members are ['ceph-1', 'ceph-2','ceph-3']
[ceph_deploy.new][DEBUG ] Monitor addrs are ['192.168.57.222','192.168.57.223', '192.168.57.224']
[ceph_deploy.new][DEBUG ] Creating a random mon key...
[ceph_deploy.new][DEBUG ] Writing monitor keyring to ceph.mon.keyring...
[ceph_deploy.new][DEBUG ] Writing initial config to ceph.conf...
此时，目录内容如下：
[root@ceph-1cluster]# ls
ceph.conf ceph-deploy-ceph.log ceph.mon.keyring
根据自己的IP配置向ceph.conf中添加public_network，并稍微增大mon之间时差允许范围(默认为0.05s，现改为2s)：
[root@ceph-1cluster]# echo public_network=192.168.57.0/24 >> ceph.conf
[root@ceph-1 cluster]# echo mon_clock_drift_allowed = 2 >> ceph.conf
[root@ceph-1 cluster]# cat ceph.conf
[global]
fsid = 0248817a-b758-4d6b-a217-11248b098e10
mon_initial_members = ceph-1, ceph-2, ceph-3
mon_host = 192.168.57.222,192.168.57.223,192.168.57.224
auth_cluster_required = cephx
auth_service_required = cephx
auth_client_required = cephx
public_network=192.168.57.0/24
mon_clock_drift_allowed = 2
开始部署monitor：
[root@ceph-1cluster]# ceph-deploy mon create-initial
..
..若干log
[root@ceph-1 cluster]# ls
ceph.bootstrap-mds.keyring ceph.bootstrap-rgw.keyring ceph.conf ceph.mon.keyring
ceph.bootstrap-osd.keyring ceph.client.admin.keyring ceph-deploy-ceph.log
查看集群状态：
[root@ceph-1cluster]# ceph -s
cluster 0248817a-b758-4d6b-a217-11248b098e10
health HEALTH_ERR
no osds
Monitorclock skew detected
monmap e1: 3 mons at{ceph-1=192.168.57.222:6789/0,ceph-2=192.168.57.223:6789/0,ceph-3=192.168.57.224:6789/0}
electionepoch 6, quorum 0,1,2 ceph-1,ceph-2,ceph-3
osdmap e1: 0 osds: 0 up, 0 in
flagssortbitwise
pgmap v2: 64 pgs, 1 pools, 0 bytes data, 0objects
0 kB used, 0kB / 0 kB avail
64 creating
开始部署OSD：
ceph-deploy--overwrite-conf osd prepare ceph-1:/dev/sdb ceph-1:/dev/sdc ceph-1:/dev/sddceph-2:/dev/sdb ceph-2:/dev/sdc ceph-2:/dev/sdd ceph-3:/dev/sdb ceph-3:/dev/sdcceph-3:/dev/sdd --zap-disk
ceph-deploy --overwrite-conf osd activate ceph-1:/dev/sdb1 ceph-1:/dev/sdc1ceph-1:/dev/sdd1 ceph-2:/dev/sdb1 ceph-2:/dev/sdc1 ceph-2:/dev/sdd1ceph-3:/dev/sdb1 ceph-3:/dev/sdc1 ceph-3:/dev/sdd1
我在部署的时候出了个小问题，有一个OSD没成功(待所有OSD部署完毕后，再重新部署问题OSD即可解决)，如果不出意外的话，集群状态应该如下：
[root@ceph-1cluster]# ceph -s
cluster 0248817a-b758-4d6b-a217-11248b098e10
health HEALTH_WARN
too few PGsper OSD (21 < min 30)
monmap e1: 3 mons at{ceph-1=192.168.57.222:6789/0,ceph-2=192.168.57.223:6789/0,ceph-3=192.168.57.224:6789/0}
electionepoch 22, quorum 0,1,2 ceph-1,ceph-2,ceph-3
osdmap e45: 9 osds: 9 up, 9 in
flagssortbitwise
pgmap v82: 64 pgs, 1 pools, 0 bytes data, 0objects
273 MB used,16335 GB / 16336 GB avail
64 active+clean
去除这个WARN，只需要增加rbd池的PG就好：
[root@ceph-1cluster]# ceph osd pool set rbd pg_num 128
set pool 0 pg_num to 128
[root@ceph-1 cluster]# ceph osd pool set rbd pgp_num 128
set pool 0 pgp_num to 128
[root@ceph-1 cluster]# ceph -s
cluster 0248817a-b758-4d6b-a217-11248b098e10
health HEALTH_ERR
19 pgs arestuck inactive for more than 300 seconds
12 pgspeering
19 pgs stuckinactive
monmap e1: 3 mons at{ceph-1=192.168.57.222:6789/0,ceph-2=192.168.57.223:6789/0,ceph-3=192.168.57.224:6789/0}
electionepoch 22, quorum 0,1,2 ceph-1,ceph-2,ceph-3
osdmap e49: 9 osds: 9 up, 9 in
flagssortbitwise
pgmap v96: 128 pgs, 1 pools, 0 bytes data, 0objects
308 MB used,18377 GB / 18378 GB avail
103 active+clean
12 peering
9 creating
4 activating
[root@ceph-1 cluster]# ceph -s
cluster 0248817a-b758-4d6b-a217-11248b098e10
health HEALTH_OK
monmap e1: 3 mons at {ceph-1=192.168.57.222:6789/0,ceph-2=192.168.57.223:6789/0,ceph-3=192.168.57.224:6789/0}
electionepoch 22, quorum 0,1,2 ceph-1,ceph-2,ceph-3
osdmap e49: 9 osds: 9 up, 9 in
flagssortbitwise
pgmap v99: 128 pgs, 1 pools, 0 bytes data, 0objects
310 MB used,18377 GB / 18378 GB avail
128 active+clean
至此，集群部署完毕。
config推送
请不要使用直接修改某个节点的/etc/ceph/ceph.conf文件的方式，而是去部署节点(此处为ceph-1:/root/cluster/ceph.conf)目录下修改。因为节点到几十个的时候，不可能一个个去修改的，采用推送的方式快捷安全！修改完毕后，执行如下指令，将conf文件推送至各个节点：
[root@ceph-1cluster]# ceph-deploy --overwrite-conf config push ceph-1 ceph-2 ceph-3
此时，需要重启各个节点的monitor服务，见下一节。
mon&osd启动方式
#monitor start/stop/restart
#ceph-1为各个monitor所在节点的主机名。
systemctl start ceph-mon@ceph-1.service
systemctl restart ceph-mon@ceph-1.service
systemctl stop ceph-mon@ceph-1.service
#OSD start/stop/restart
#0为该节点的OSD的id，可以通过`ceph osd tree`查看
systemctl start/stop/restart ceph-osd@0.service
[root@ceph-1 cluster]# ceph osd tree
ID WEIGHT TYPE NAME UP/DOWNREWEIGHT PRIMARY-AFFINITY
-1 17.94685 rootdefault
-2 5.98228 hostceph-1
0 1.99409 osd.0 up 1.00000 1.00000
1 1.99409 osd.1 up 1.00000 1.00000
8 1.99409 osd.2 up 1.00000 1.00000
-3 5.98228 hostceph-2
2 1.99409 osd.3 up 1.00000 1.00000
3 1.99409 osd.4 up 1.00000 1.00000
4 1.99409 osd.5 up 1.00000 1.00000
-4 5.98228 hostceph-3
5 1.99409 osd.6 up 1.00000 1.00000
6 1.99409 osd.7 up 1.00000 1.00000
7 1.99409 osd.8 up 1.00000 1.00000
原文来自：TStack 公众号
