[自动化]基于kolla的自动化部署ceph集群
阅读原文时间:2022年03月01日阅读:1
kolla-ceph来源:

项目中的部分代码来自于kolla和kolla-ansible

kolla-ceph的介绍:

1、镜像的构建很方便, 基于容器的方式部署,创建、删除方便

2、kolla-ceph的操作幂等,多次执行不会产生副作用

3、使用kolla-ceph(基于ansible)流程化部署

4、通过给磁盘打上相应标签,创建osd非常简单

5、升级便捷,通过构建新的ceph镜像,upgrade既可

6、自动根据osd节点数量来设置故障域: "osd" 或 "host",及配置对应的副本数

项目结构

Auto_Ceph
├── 00-hosts
├── README.md
├── action_plugins
├── action.yml
├── bin
├── build
├── config
├── group_vars
├── library
├── os.yml
├── requirements.txt
├── roles
└── site.yml

点击查看项目:kolla-ceph自动化部署ceph集群

系统: Centos

环境: 3台虚拟机(可采用单节点或多节点),下载Auto_Ceph项目放在/root/目录下

ceph集群节点规划、网络规划

vi /root/Auto_Ceph/00-hosts

# storage_interface=eth0 ceph集群管理接口,必须配置
# cluster_interface=eth1 ceph集群数据同步接口,为空默认就是storage_interface

[all:vars]
storage_interface=eth0
cluster_interface=eth1

[mon]
172.20.163.244  # 同时是部署节点
172.20.163.67
172.20.163.238

[mgr]
172.20.163.244
172.20.163.67
172.20.163.238 

[osd]
172.20.163.244
172.20.163.67
172.20.163.238 

[rgw]
172.20.163.244
172.20.163.67
172.20.163.238 

[mds]

下载ceph镜像<部署节点操作>

部署节点:可以是任意一台mon组的节点(172.20.163.244)

  1. 在线部署: 下载ceph镜像、安装ansible、kolla-ceph<部署节点操作>

    1. type wget || yum install wget -y

    2. wget https://bootstrap.pypa.io/pip/2.7/get-pip.py --no-check-certificate

    3. python get-pip.py

    4.ceph镜像下载安装依赖
    pip install -r /root/Auto_Ceph/requirements.txt --ignore-installed

    当出现以下报错的时候执行:pip install GitPython, 再执行pip install kolla==9.4.0
    ERROR: pip's legacy dependency resolver does not consider dependency conflicts when selecting packages. This behaviour is the source of the following dependency conflicts.
    gitdb2 4.0.2 requires gitdb>=4.0.1, but you'll have gitdb 0.6.4 which is incompatible.
    gitpython 2.1.15 requires gitdb2<3,>=2, but you'll have gitdb2 4.0.2 which is incompatible.

    1. 下载docker
      sh /root/Auto_Ceph/bin/install -D

    2. 运行docker
      sh /root/Auto_Ceph/bin/install -I
      systemctl status docker

    3. 下载registry
      sh /root/Auto_Ceph/bin/install -R

    4. 运行registry
      docker run -d -v /opt/registry:/var/lib/registry -p 5000:5000 --restart=always --name registry registry:latest

    5. 修改配置
      vi /root/Auto_Ceph/build/ceph-build.conf
      registry = 172.17.2.179:5000 # 必须按照实际修改, 其它默认既可

    6. 开始构建ceph镜像, 查看镜像
      cd /root/Auto_Ceph/build/ && sh build.sh --tag nautilus
      docker image ls
      REPOSITORY TAG IMAGE ID CREATED SIZE
      172.20.163.77:5000/kolla-ceph/centos-binary-ceph-mon nautilus a5e8a5ff08fc 13 days ago 792MB
      172.20.163.77:5000/kolla-ceph/centos-binary-ceph-osd nautilus 118b704bcf88 13 days ago 793MB
      172.20.163.77:5000/kolla-ceph/centos-binary-cephfs-fuse nautilus 6b00fc4b6e2e 13 days ago 792MB
      172.20.163.77:5000/kolla-ceph/centos-binary-ceph-mds nautilus b206c578e594 13 days ago 792MB
      172.20.163.77:5000/kolla-ceph/centos-binary-ceph-rgw nautilus e9f5e4bca8ab 13 days ago 792MB
      172.20.163.77:5000/kolla-ceph/centos-binary-ceph-mgr nautilus b561bf427142 13 days ago 792MB
      172.20.163.77:5000/kolla-ceph/centos-binary-ceph-base nautilus eae0898ce208 13 days ago 792MB
      172.20.163.77:5000/kolla-ceph/centos-binary-base nautilus d48db6e179f9 13 days ago 410MB

    7. type ansible || yum install ansible -y

    8. 部署节点与节点之间ssh免密配置

    9. 安装kolla-ceph工具
      cd /root/Auto_Ceph/bin && sh install -K

  2. 在线部署: 下载docker<除部署节点外, 在其它节点操作以下步骤>

    1. type wget || yum install wget -y

    2. wget https://bootstrap.pypa.io/pip/2.7/get-pip.py --no-check-certificate

    3. python get-pip.py

    4. 安装docker模块
      pip install docker

    5. 安装并运行docker
      scp /root/Auto_Ceph/bin/install ${target_host}:/root/
      sh /root/install -D && sh /root/install -I

部署ceph集群

1. 修改参数<部署节点操作>
vi /root/Auto_Ceph/config/globals.yml 

   ceph_tag: "nautilus"
   docker_registry: "仓库地址:端口"
   ceph_osd_store_type: "bluestore"
   ceph_pool_pg_num: 32 # 设置你的pg数
   ceph_pool_pgp_num: 32 # 设置你的pgp数
   enable_ceph_rgw: "true or false"
   enable_ceph_mds: "true or false"
2. kolla-ceph部署使用<部署节点操作>
2.1 初始化ceph主机节点

    kolla-ceph -i /root/Auto_Ceph/00-hosts os

2.2 部署前检查配置

    kolla-ceph -i /root/Auto_Ceph/00-hosts prechecks

2.3 部署ceph集群

    1、bluestore osd: 为每个osd节点的磁盘打上标签
       parted  /dev/vdc  -s  -- mklabel  gpt  mkpart KOLLA_CEPH_OSD_BOOTSTRAP_BS  1 -1
    2、部署ceph-mon、ceph-osd、ceph-mgr、ceph-rgw、ceph-mds
       kolla-ceph -i /root/Auto_Ceph/00-hosts deploy
    3、docker exec ceph_mon ceph -s
         cluster:
           id:     4a9e463a-4853-4237-a5c5-9ae9d25bacda
           health: HEALTH_OK

         services:
           mon: 3 daemons, quorum 172.20.163.67,172.20.163.77,172.20.163.238 (age 2h)
           mgr: 172.20.163.238(active, since 2h), standbys: 172.20.163.77, 172.20.163.67
           mds: cephfs:1 {0=devops2=up:active} 2 up:standby
           osd: 4 osds: 4 up (since 2h), 4 in (since 13d)
           rgw: 1 daemon active (radosgw.gateway)

         data:
           pools:   7 pools, 104 pgs
           objects: 260 objects, 7.6 KiB
           usage:   4.1 GiB used, 76 GiB / 80 GiB avail
           pgs:     104 active+clean   

2.4 删除操作: ceph集群容器和volume

    kolla-ceph -i /root/Auto_Ceph/00-hosts  destroy --yes-i-really-really-mean-it

2.5 升级操作

    1、cd /root/Auto_Ceph/build/ && sh build.sh --tag new_ceph_version
    2、修改最新ceph_tag: "new_ceph_version"
    3、kolla-ceph -i /root/Auto_Ceph/00-hosts upgrade

2.6 单独更换部署osd

    kolla-ceph -i /root/Auto_Ceph/00-hosts -t ceph-osd

2.7 开启ceph dashborad
    enable_ceph_dashboard: true
    kolla-ceph -i /root/Auto_Ceph/00-hosts start-dashborad

2.8 启用对象网关管理前端
    enable_ceph_rgw: true
    kolla-ceph -i /root/Auto_Ceph/00-hosts start-rgw-front
3. 磁盘打标签介绍
3.1. bluestore wal db共用一块盘打标签方式
  1. parted /dev/vdc -s -- mklabel gpt mkpart KOLLA_CEPH_OSD_BOOTSTRAP_BS 1 -1
3.2. bluestore 分离db和wal打标签方式

为了提高 ceph 性能,且ssd磁盘数量有限,通常将db和wal存放在单独的 ssd 磁盘上

  # SSD磁盘:vdb vdd HDD磁盘:vdc
    1. 指定元数据分区
       parted /dev/vdc -s -- mklabel  gpt mkpart KOLLA_CEPH_OSD_BOOTSTRAP_BS_BLUE1 1 100
    2. 指定block 分区
       parted /dev/vdc -s -- mkpart KOLLA_CEPH_OSD_BOOTSTRAP_BS_BLUE1_B 101 100%

    3. 指定block.wal分区
       parted /dev/vdb -s -- mklabel  gpt mkpart KOLLA_CEPH_OSD_BOOTSTRAP_BS_BLUE1_W 1 1000
    4. 指定block.db分区
       parted /dev/vdd -s -- mklabel  gpt mkpart KOLLA_CEPH_OSD_BOOTSTRAP_BS_BLUE1_D 1 10000

block.db 分区的大小为 block 分区 的 4%大小

3.3 filestore 打标签方式
  1. parted /dev/vdc -s -- mklabel gpt mkpart KOLLA_CEPH_OSD_BOOTSTRAP 1 -1
3.4 filestore 指定日志单独分区

filestore 为了提高 ceph 性能,通常将日志存放在单独的 ssd 磁盘上

# SSD磁盘:vdb    HDD磁盘:vdc vdd
1. vdc 作为数据盘
   parted /dev/vdc -s -- mklabel gpt mkpart KOLLA_CEPH_OSD_BOOTSTRAP_FILE1 1 -1
2. vdd 作为数据盘
   parted /dev/vdd -s -- mklabel gpt mkpart KOLLA_CEPH_OSD_BOOTSTRAP_FILE2 1 -1
3. vdb作为vdc、vdd 的journal盘
   parted /dev/vdb -s -- mklabel gpt
   parted /dev/vdb -s -- mkpart KOLLA_CEPH_OSD_BOOTSTRAP_FILE1_J 4M 2G
   parted /dev/vdb -s -- mkpart KOLLA_CEPH_OSD_BOOTSTRAP_FILE2_J 2G 4G

运维操作

1、任意一台montior节点进入ceph环境. 既可以正常执行ceph命令运维操作
   docker exec -it ceph_mon bash
   ceph -s

2、或者直接外部操作
   docker exec ceph_mon ceph -s

3、osd故障操作
   docker exec ceph_mon ceph osd crush rm osd.1
   docker exec ceph_mon ceph osd auth rm osd.1
   docker exec ceph_mon ceph osd rm osd.1
   到故障osd节点把容器给干掉,然后换新盘: docker rm -f ceph_osd_1
   为新磁盘盘打标签: parted  /dev/vdc  -s  -- mklabel  gpt  mkpart KOLLA_CEPH_OSD_BOOTSTRAP_BS  1 -1
   部署新osd: kolla-ceph -i /root/Auto_Ceph/00-hosts -t ceph-osd

ceph dashboard

ceph 集群状态

ceph rgw