Ceph集群生产环境安装部署

我会带着你远行 2022-05-13 07:28 509阅读 0赞

## 前言 ##

ceph的组件以及工作流程非常的复杂,是一个庞大的系统,在尝试ceph之前尽量多查阅官方的文档,理解ceph的mon/osd/mds/pg/pool等各组件/Unit的协同工作方式  
[**Ceph官方文档**][Ceph]

## 一、配置规划： ##

![这里写图片描述][70]

## 二、部署 ##

**1. ntp-server开启ntp服务：**

apt-get install ntp ntpdate ntp-doc
    systemctl enable ntp
    systemctl start ntp

**2. ceph node**

**三台node全部执行如下操作：**  
磁盘分区规划如顶部表格，按照规划写的磁盘划分脚本，分别在3台node上执行脚本：

# cat ~/parted.sh 
    
    #!/bin/bash
    set -e
    if [ ! -x "/usr/sbin/parted" ]; then
    echo "This script requires /sbin/parted to run!" >&2
    exit 1
    fi
    DISKS="d e f g h i j k l m n o p"
    for i in ${DISKS}; do
      echo "Creating partitions on /dev/sd${i} ..."
      parted -a optimal --script /dev/sd${i} -- mktable gpt
      parted -a optimal --script /dev/sd${i} -- mkpart primary xfs 0% 100%
      sleep 1
      #echo "Formatting /dev/sd${i}1 ..."
      mkfs.xfs -f /dev/sd${i}1 &
    done
    
    SSDS="b c"
    for i in ${SSDS}; do
      parted -s /dev/sd${i} mklabel gpt
      parted -s /dev/sd${i} mkpart primary 0% 10%
      parted -s /dev/sd${i} mkpart primary 11% 20%
      parted -s /dev/sd${i} mkpart primary 21% 30%
      parted -s /dev/sd${i} mkpart primary 31% 40%
      parted -s /dev/sd${i} mkpart primary 41% 50%
      parted -s /dev/sd${i} mkpart primary 51% 60%
      parted -s /dev/sd${i} mkpart primary 61% 70%
    done
    
    chown -R ceph:ceph /dev/sdb[1-7]
    chown -R ceph:ceph /dev/sdc[1-7]

添加 /etc/hosts解析，并scp到3台node上:

192.168.20.112  h020112
    192.168.20.113  h020113
    192.168.20.114  h020114

关闭防火墙、selinux，添加定时同步时间计划任务：

sed -i 's/SELINUX=.*/SELINUX=disabled/' /etc/selinux/config
     setenforce 0
     systemctl stop firewalld
     systemctl disable firewalld
     
     cat >>/etc/crontab<<EOF
     × */1 * * * root ntpdate 192.168.9.28 && --systohc
     EOF

cd /etc/yum.repos.d
    mv CentOS-Base.repo CentOS-Base.repo.bak
    wget -O /etc/yum.repos.d/CentOS-Base.repo http://mirrors.aliyun.com/repo/Centos-7.repo
    yum -y install ntp ntpdate
    yum -y install ceph

**在node1(192.168.20.112)上执行:**

wget https://download.ceph.com/rpm-kraken/el7/noarch/ceph-deploy-1.5.37-0.noarch.rpm
    yum -y install ceph-deploy-1.5.37-0.noarch.rpm
    
    # 创建秘钥，传送到各节点，实现无秘钥登录
    ssh-keygen
    ssh-copy-id h020112
    ssh-copy-id h020113
    ssh-copy-id h020114
    
    # 新建集群，生成配置文件
    mkdir ceph-cluster && cd ceph-cluster
    ceph-deploy new h020112 h020113 h020114
    
    ## 修改默认生成的ceph.conf，增加如下配置段：
    # 80G日志盘
    osd_journal_size = 81920 
    public_network= 192.168.20.0/24 
    # 副本pg数为2，默认为3，最小工作size为默认size - (默认size/2) 
    osd pool default size = 2
    # 官方建议平均每个osd 的pg数量不小于30，即pg num > (osd_num) * 30 / 2(副本数)
    osd pool default pg num = 1024
    osd pool default pgp num = 1024
    
    # 传送ceph.conf
    ceph-deploy --overwrite-conf config push h020112 h020113 h020114
    
    # 若有RuntimeError: bootstrap-mds keyring not found; run 'gatherkeys'报错，则执行如下命令传送key
    ceph-deploy gatherkeys yh020112 h020113 h020114
    
    # 初始化mon节点
    ceph-deploy mon create-initial
    ceph -s # 查看mon是否添加成功
    
    # ceph 集群加入OSD
    cd /etc/ceph # 一定要进入这个目录下
    
    # 执行如下脚本，日志盘和数据盘的对应关系要再三确认
    [root@h20112 ceph]# cat add_osd.sh 
    #!/bin/bash
    
    for ip in $(cat ~/ceph-cluster/cephosd.txt)  
    do
    echo ----$ip-----------;
    ceph-deploy --overwrite-conf osd prepare $ip:sdd1:/dev/sdb1 $ip:sde1:/dev/sdb2 $ip:sdf1:/dev/sdb3 $ip:sdg1:/dev/sdb4  $ip:sdh1:/dev/sdb5  $ip:sdi1:/dev/sdb6  \
       $ip:sdj1:/dev/sdc1 $ip:sdk1:/dev/sdc2 $ip:sdl1:/dev/sdc3 $ip:sdm1:/dev/sdc4  $ip:sdn1:/dev/sdc5  $ip:sdo1:/dev/sdc6  $ip:sdp1:/dev/sdc7
    done
    
    
    for ip in $(cat ~/ceph-cluster/cephosd.txt)
    do
    echo ----$ip-----------;
    ceph-deploy osd activate $ip:sdd1:/dev/sdb1 $ip:sde1:/dev/sdb2 $ip:sdf1:/dev/sdb3 $ip:sdg1:/dev/sdb4  $ip:sdh1:/dev/sdb5  $ip:sdi1:/dev/sdb6  \
       $ip:sdj1:/dev/sdc1 $ip:sdk1:/dev/sdc2 $ip:sdl1:/dev/sdc3 $ip:sdm1:/dev/sdc4  $ip:sdn1:/dev/sdc5  $ip:sdo1:/dev/sdc6  $ip:sdp1:/dev/sdc7
    done

查看结果：

[root@h20112 ceph]# ceph osd tree
    ID WEIGHT   TYPE NAME           UP/DOWN REWEIGHT PRIMARY-AFFINITY 
    -1 34.04288 root default                                          
    -2 11.34763     host h020112                                   
     0  0.87289         osd.0            up  1.00000          1.00000 
     3  0.87289         osd.3            up  1.00000          1.00000 
     4  0.87289         osd.4            up  1.00000          1.00000 
     5  0.87289         osd.5            up  1.00000          1.00000 
     6  0.87289         osd.6            up  1.00000          1.00000 
     7  0.87289         osd.7            up  1.00000          1.00000 
     8  0.87289         osd.8            up  1.00000          1.00000 
     9  0.87289         osd.9            up  1.00000          1.00000 
    10  0.87289         osd.10           up  1.00000          1.00000 
    11  0.87289         osd.11           up  1.00000          1.00000 
    12  0.87289         osd.12           up  1.00000          1.00000 
    13  0.87289         osd.13           up  1.00000          1.00000 
    14  0.87289         osd.14           up  1.00000          1.00000 
    -3 11.34763     host h020113                                   
     1  0.87289         osd.1            up  1.00000          1.00000 
    15  0.87289         osd.15           up  1.00000          1.00000 
    16  0.87289         osd.16           up  1.00000          1.00000 
    17  0.87289         osd.17           up  1.00000          1.00000 
    18  0.87289         osd.18           up  1.00000          1.00000 
    19  0.87289         osd.19           up  1.00000          1.00000 
    20  0.87289         osd.20           up  1.00000          1.00000 
    21  0.87289         osd.21           up  1.00000          1.00000 
    22  0.87289         osd.22           up  1.00000          1.00000 
    23  0.87289         osd.23           up  1.00000          1.00000 
    24  0.87289         osd.24           up  1.00000          1.00000 
    25  0.87289         osd.25           up  1.00000          1.00000 
    26  0.87289         osd.26           up  1.00000          1.00000 
    -4 11.34763     host h020114                                   
     2  0.87289         osd.2            up  1.00000          1.00000 
    27  0.87289         osd.27           up  1.00000          1.00000 
    28  0.87289         osd.28           up  1.00000          1.00000 
    29  0.87289         osd.29           up  1.00000          1.00000 
    30  0.87289         osd.30           up  1.00000          1.00000 
    31  0.87289         osd.31           up  1.00000          1.00000 
    32  0.87289         osd.32           up  1.00000          1.00000 
    33  0.87289         osd.33           up  1.00000          1.00000 
    34  0.87289         osd.34           up  1.00000          1.00000 
    35  0.87289         osd.35           up  1.00000          1.00000 
    36  0.87289         osd.36           up  1.00000          1.00000 
    37  0.87289         osd.37           up  1.00000          1.00000 
    38  0.87289         osd.38           up  1.00000          1.00000 
    
    [root@h20112 ceph]# ceph -s
        cluster 6661d89d-5895-4bcb-9b11-9400638afc85
         health HEALTH_OK
         monmap e1: 3 mons at {h020112=192.168.20.112:6789/0,h020113=192.168.20.113:6789/0,h020114=192.168.20.114:6789/0}
                election epoch 6, quorum 0,1,2 h020112,h020113,h020114
         osdmap e199: 39 osds: 39 up, 39 in
                flags sortbitwise,require_jewel_osds
          pgmap v497: 1024 pgs, 1 pools, 0 bytes data, 0 objects
                4385 MB used, 34854 GB / 34858 GB avail
                    1024 active+clean
    
    # 若配置文件里指定的pg_num 和 php_num未生效，使用命令指定：
    sudo ceph osd pool set rbd pg_num 1024
    sudo ceph osd pool set rbd pgp_num 1024

## 自定义crush分布式调度规则： ##

ceph一共有如下层级的管理单位，从上到下层级依次提升，可以灵活地按照物理的逻辑粒度，将osd关联到不同的主机、机位、机架、pdu、机房、区域等管理单位，每一层级的整体权重值等于该层级下所有OSD的权重之和。

type 0 osd
    type 1 host
    type 2 chassis
    type 3 rack
    type 4 row
    type 5 pdu
    type 6 pod
    type 7 room
    type 8 datacenter
    type 9 region
    type 10 root

**自定义crush视图:**  
这里只写操作方法，不作实施，本次操作的环境硬盘数目、硬盘规格、主机位置均一致，暂时不作crush调整

ceph osd crush add-bucket hnc root #添加root层级名为hnc的bucket
    ceph osd crush add-bucket rack0 rack  #添加rack层级名为rack0的bucket
    ceph osd crush add-bucket rack1 rack  #添加rack层级名为rack1的bucket
    ceph osd crush add-bucket rack2 rack  #添加rack层级名为rack2的bucket
    ceph osd crush move rack0 root=hnc  #将rack0移入hnc之下
    ceph osd crush move rack0 root=hnc  #将rack1移入hnc之下
    ceph osd crush move rack0 root=hnc  #将rack2移入hnc之下
    ceph ods crush move h020112 rack=rack0  #将host h020112移入rack0层级下
    ceph ods crush move h020113 rack=rack1  #将host h020113移入rack1层级下
    ceph ods crush move h020114 rack=rack2  #将host h020114移入rack2层级下
    
    ceph osd getcrushmap -o map_old    导出map
    crushtool -d map_old -o map_old.txt  转化成可编辑格式
    
    crushtool -c map_new.txt -o map_new  还原为map
    ceph osd setcrushmap -i map_new     将map导入ceph

修改配置文件,防止ceph自动更新crushmap

echo ‘osd_crush_update_on_start = false‘ >> /etc/ceph/ceph.conf
    /etc/init.d/ceph restart

使用如下样例crash（map\_new.txt）配置:

--------------------------------------
    # begin crush map
    tunable choose_local_tries 0
    tunable choose_local_fallback_tries 0
    tunable choose_total_tries 50
    tunable chooseleaf_descend_once 1
    tunable straw_calc_version 1
    
    # devices
    device 0 osd.0
    device 1 osd.1
    device 2 osd.2
    device 3 osd.3
    device 4 osd.4
    device 5 osd.5
    device 6 osd.6
    device 7 osd.7
    device 8 osd.8
    device 9 osd.9
    device 10 osd.10
    device 11 osd.11
    
    # types
    type 0 osd
    type 1 host
    type 2 chassis
    type 3 rack
    type 4 row
    type 5 pdu
    type 6 pod
    type 7 room
    type 8 datacenter
    type 9 region
    type 10 root
    
    # buckets
    root default {
    	id -1		# do not change unnecessarily
    	# weight 0.000
    	alg straw
    	hash 0	# rjenkins1
    }
    
    # 权重值一般根据磁盘容量(基数)与性能(倍率)调整，例如设置1T为1.00，2T为2.00，hdd倍数为1，ssd倍数为2
    host h020112 {
    	id -2		# do not change unnecessarily
    	# weight 0.000
    	alg straw
    	hash 0	# rjenkins1
    	item osd.0 weight 2.000
    	item osd.1 weight 2.000
    	item osd.2 weight 2.000
    	item osd.3 weight 2.000
    	# 需写出全部osd，这里省略不写这么多了
    }
    host h020113 {
    	id -3		# do not change unnecessarily
    	# weight 0.000
    	alg straw
    	hash 0	# rjenkins1
    	item osd.4 weight 2.000
    	item osd.5 weight 2.000
    	item osd.6 weight 2.000
    	item osd.7 weight 2.000
    }
    host h020114 {
    	id -4		# do not change unnecessarily
    	# weight 0.000
    	alg straw
    	hash 0	# rjenkins1
    	item osd.8 weight  2.000 
    	item osd.9 weight  2.000  
    	item osd.10 weight 2.000  
    	item osd.11 weight 2.000  
    }
    rack rack0 {
    	id -6		# do not change unnecessarily
    	# weight 0.000
    	alg straw
    	hash 0	# rjenkins1
    	item h020112 weight 8.000
    }
    rack rack1 {
    	id -7		# do not change unnecessarily
    	# weight 0.000
    	alg straw
    	hash 0	# rjenkins1
    	item h020113 weight 8.000
    }
    rack rack2 {
    	id -8		# do not change unnecessarily
    	# weight 0.000
    	alg straw
    	hash 0	# rjenkins1
    	item h020114 weight 8.000
    }
    root hnc {
    	id -5		# do not change unnecessarily
    	# weight 0.000
    	alg straw
    	hash 0	# rjenkins1
    	item rack0 weight 8.000
    	item rack1 weight 8.000
    	item rack2 weight 8.000
    }
    
    # rules
    rule replicated_ruleset {                         #规则集的命名，创建pool时可以指定rule集
    	ruleset 0                                     #rules集的编号，顺序编即可
    	type replicated                               #定义pool类型为replicated(还有esurecode模式)
    	min_size 1                                    #pool中最小指定的副本数量不能小1
    	max_size 10                                   #pool中最大指定的副本数量不能大于10
    	step take hnc                             #定义pg查找副本的入口点
    	step chooseleaf firstn 0 type rack            #选叶子节点、深度优先、隔离rack
    	step emit
    }
    # end crush map

将修改后的crushmap编译并且注入集群中

crushtool -c map_new.txt -o map_new
    ceph osd setcrushmap -i map_new
    ceph osd tree
    ceph osd crush rm default  # 删除默认crush map
    systemctl stop ceph\*.service ceph\*.target  #关闭所有服务
    systemctl start ceph.target   #启动服务

[Ceph]: http://docs.ceph.com/docs/master/
[70]: /images/20220513/1541a18ac24a44249d3ce5514a1a7b7e.png