ceph集群安装（块设备）3PC（3mon+3osd+3mgr）+1ceph-deploy机器

ceph节点操作：

创建sudo用户，加入公钥，允许部署节点访问。
disable /etc/selinux/config
setenforce 0
yum install yum-plugin-priorities -y
时间同步：
chronyc makestep

部署节点操作：

yum install -y https://dl.fedoraproject.org/pub/epel/epel-release-latest-7.noarch.rpm

cat << EOM > /etc/yum.repos.d/ceph.repo
[ceph-noarch]
name=Ceph noarch packages
baseurl=https://download.ceph.com/rpm-mimic/el7/noarch
enabled=1
gpgcheck=1
type=rpm-md
gpgkey=https://download.ceph.com/keys/release.asc
EOM

cat << EOM >>/etc/hosts
192.168.200.237 ceph-1
192.168.200.238 ceph-2
192.168.200.239 ceph-3
EOM

创建部署配置文件等使用的目录，所有操作在该目录下执行：
mkdir my-cluster
cd my-cluster

清理旧的ceph：
ceph-deploy purge {ceph-node} [{ceph-node}]
ceph-deploy purgedata {ceph-node} [{ceph-node}]
ceph-deploy forgetkeys
rm ceph.*

开始部署：

ceph-deploy --username test new ceph-1 ceph-2 ceph-3

vi ceph.conf
public network = 192.168.200.0/24
osd pool default pg num = 128
osd pool default pgp num = 128
[mon]
mon allow pool delete = true

export CEPH_DEPLOY_REPO_URL=http://mirrors.163.com/ceph/rpm-mimic/el7
export CEPH_DEPLOY_GPG_URL=http://mirrors.163.com/ceph/keys/release.asc
安装ceph
ceph-deploy --username test install ceph-1 ceph-2 ceph-3
安装mon
ceph-deploy --username test mon create ceph-1 ceph-2 ceph-3
ceph-deploy --username test mon create-initial
同步配置
ceph-deploy --username test admin ceph-1 ceph-2 ceph-3
安装mgr
ceph-deploy --username test mgr ceph-1 ceph-2 ceph-3
检查ceph osd使用的磁盘
ceph-deploy disk list ceph-1 ceph-2 ceph-3
擦净磁盘
ceph-deploy disk zap ceph-1 /dev/sdb
安装osd
ceph-deploy --username test osd create --data /dev/sdb ceph-1
ceph-deploy --username test osd create --data /dev/sdb ceph-2
ceph-deploy --username test osd create --data /dev/sdb ceph-3
ceph-deploy --username test osd create --data /dev/sdc ceph-1
ceph-deploy --username test osd create --data /dev/sdc ceph-2
ceph-deploy --username test osd create --data /dev/sdc ceph-3

#安装完成。

pg计算

根据官网计算方法http://ceph.com/pgcalc/ 
Total PGs = ((Total_number_of_OSD * 100) / max_replication_count) / pool_count
osd:6   relications: 3    pool: 2
pg=100, 接近128 , 所以pg_num 设为128

重新推送配置：
ceph-deploy --username test --overwrite-conf config push ceph-1 ceph-2 ceph-3
推送后ceph节点重启操作：
systemctl restart ceph-mon.target

==============================================

ceph节点操作：

ceph状态检查：

ceph health detail
ceph -s
检查mon法定人数状态
ceph quorum_status --format json-pretty

1. ceph CRUSH map 增加规则

手动处理规则方式：

创建root，host
ceph osd crush add-bucket root-nvme root
ceph osd crush add-bucket root-ssd root
ceph osd crush add-bucket host1-nvme host
ceph osd crush add-bucket host2-nvme host
ceph osd crush add-bucket host3-nvme host
ceph osd crush add-bucket host1-ssd host
ceph osd crush add-bucket host2-ssd host
ceph osd crush add-bucket host3-ssd host
把host加入root
ceph osd crush move host1-ssd root=root-ssd
ceph osd crush move host2-ssd root=root-ssd
ceph osd crush move host3-ssd root=root-ssd
ceph osd crush move host3-nvme root=root-nvme
ceph osd crush move host2-nvme root=root-nvme
ceph osd crush move host1-nvme root=root-nvme
把osd加入host
ceph osd crush move osd.0 host=host1-nvme
ceph osd crush move osd.1 host=host2-nvme
ceph osd crush move osd.2 host=host3-nvme
ceph osd crush move osd.3 host=host1-ssd
ceph osd crush move osd.4 host=host2-ssd
ceph osd crush move osd.5 host=host3-ssd

导出CRUSH map：
ceph osd getcrushmap -o crushmap.txt
crushtool -d crushmap.txt -o crushmap-decompile

修改规则vi crushmap-decompile：
# rules
rule nvme {
        id 1
        type replicated
        min_size 1
        max_size 10
        step take root-nvme
        step chooseleaf firstn 0 type host
        step emit
}

rule ssd {
        id 2
        type replicated
        min_size 1
        max_size 10
        step take root-ssd
        step chooseleaf firstn 0 type host
        step emit
}

导入CRUSH map：
crushtool -c crushmap-decompile -o crushmap-compiled
ceph osd setcrushmap -i crushmap-compiled

配置ceph.conf，让OSD启动不更新crushmap
[osd]
osd crush update on start = false

自动处理规则方式（磁盘智能分组）：

ceph中的每个osd设备都可以选择一个class类型与之关联，默认情况下，在创建osd的时候会自动识别设备类型，然后设置该设备为相应的类。通常有三种class类型：hdd，ssd，nvme。

###手动修改class标签
#查看当前集群布局
ceph osd tree
#查看crush class
ceph osd crush class ls
#删除osd.0，osd.1，osd.2的class
for i in 0 1 2;do ceph osd crush rm-device-class osd.$i;done
#设置osd.0，osd.1，osd.2的class为nvme
for i in 0 1 2;do ceph osd crush set-device-class nvme osd.$i;done
#创建一个优先使用nvme设备的crush规则
ceph osd crush rule create-replicated rule-auto-nvme default host nvme
#创建一个优先使用ssd设备的crush规则
ceph osd crush rule create-replicated rule-auto-ssd default host ssd
#查看集群的rule
ceph osd crush rule ls

2. 创建一个新的pool

#创建
ceph osd pool create pool-ssd 128
#创建并设置规则：
ceph osd pool create pool-ssd 128 128 rule-auto-ssd
# 设置pool的类型
ceph osd pool application enable pool-ssd rbd

osd pool 状态检查

#查看与设置pg num
ceph osd pool get hdd_pool pg_num
ceph osd pool get hdd_pool pgp_num
ceph osd pool set rbd pg_num 256
ceph osd pool set rbd pgp_num 256
#ods与pool状态
ceph osd tree
rados df
ceph df
ceph osd lspools

#查看crush_rule及pool的rule
ceph osd crush rule ls
ceph osd pool get pool-ssd crush_rule
#查看pool详情
ceph osd pool ls detail

3. 创建对象测试pool

rados -p pool-ssd ls
echo "hahah" >test.txt
rados -p pool-ssd put test test.txt 
rados -p pool-ssd ls
#查看该对象的osd组
ceph osd map pool-ssd test
#删除
rados rm -p pool-ssd test

其它

#删除pool
ceph osd pool delete pool-ssd pool-ssd --yes-i-really-really-mean-it
#设置 pool的crush_rule
ceph osd pool set pool-ssd crush_rule rule-ssd

客户端使用块设备

给需要使用ceph存储的客户端部署ceph

部署节点操作：

export CEPH_DEPLOY_REPO_URL=http://mirrors.163.com/ceph/rpm-mimic/el7
export CEPH_DEPLOY_GPG_URL=http://mirrors.163.com/ceph/keys/release.asc
ceph-deploy install ceph-client
ceph-deploy admin ceph-client

ceph客户端操作：

#创建一个块设备image
rbd create pool-ssd/foo --size 1024 --image-feature layering
#把 image 映射为块设备
rbd map pool-ssd/foo --name client.admin

mkfs.xfs /dev/rbd0
mount /dev/rbd0 /mnt/rbdtest/

块设备查询

#ceph客户端操作
镜像信息
rbd ls pool-ssd
rbd info pool-ssd/foo
查看已映射块设备
rbd showmapped
取消块设备映射
rbd unmap /dev/rbd0
删除块设备映像
rbd rm pool-ssd/foo