Ceph的基本安装

birk

　　一、基本环境介绍：
　　  本文采用ceph-deploy工具进行ceph安装，ceph-deploy可以单独作为一个admin节点，也可以装在任意node节点上。

　　系统环境如下：
　　1、  系统采用redhat-6.5x86_64 basic_server安装，总共3个节点,ntp时间同步
　　2、  关闭selinux，用epel及ceph官方源，版本为0.86
　　3、  3个节点已做互信，并配置host，每个节点上有3个磁盘做osd
　　4、  升级内核为3.18，并从新内核启动。
　　

　　二、安装步骤：
　　1、  每个node上设置iptables规则或关闭iptables(eth0指ceph网络所在的网卡名称)
iptables -A INPUT -i eth0 -p tcp -s0.0.0.0/0 --dport 6789 -j ACCEPT
iptables -A INPUT -i eth0 -m multiport -ptcp -s 0.0.0.0/0 --dports 6800:7300 -j ACCEPT
service iptables save　　2、  格式化、挂载osd
yum -y install xfsprogs
mkfs.xfs /dev/sdb
mkdir /osd{0..2}
#blkid 查看sdb的UUID
echo 'UUID=89048e27-ff01-4365-a103-22e95fb2cc93/osd0 xfs    noatime,nobarrier,nodiratime    00' >> /etc/fstab　　一个磁盘对应一个osd，每个节点都创建osd0、osd1、osd2目录，对应的磁盘挂载相应目录即可。
　　3、安装ceph部署工具
#mkdir ceph  #最好创建一个目录，因为安装ceph过程中，会在安装目录生成一些文件                  
#cd ceph
#yum -y install ceph-deploy　　4、创建mon
ceph-deploy new node1 nod2 node3 #这个命令其实就是仅仅生成了ceph.conf和ceph.mon.keyring 文件
vim ceph.conf 追加以下内容(根据需求更改)
debug_ms = 0
mon_clock_drift_allowed = 1
osd_pool_default_size = 2    #副本数量
osd_pool_default_min_size = 1
osd_pool_default_pg_num = 128  #pg数量
osd_pool_default_pgp_num = 128
osd_crush_chooseleaf_type = 0
debug_auth = 0/0
debug_optracker = 0/0
debug_monc = 0/0
debug_crush = 0/0
debug_buffer = 0/0
debug_tp = 0/0
debug_journaler = 0/0
debug_journal = 0/0
debug_lockdep = 0/0
debug_objclass = 0/0
debug_perfcounter = 0/0
debug_timer = 0/0
debug_filestore = 0/0
debug_context = 0/0
debug_finisher = 0/0
debug_heartbeatmap = 0/0
debug_asok = 0/0
debug_throttle = 0/0
debug_osd = 0/0
debug_rgw = 0/0
debug_mon = 0/0
osd_max_backfills = 4
filestore_split_multiple = 8
filestore_fd_cache_size = 1024
filestore_queue_committing_max_bytes =1048576000
filestore_queue_max_ops = 500000
filestore_queue_max_bytes = 1048576000
filestore_queue_committing_max_ops = 500000
osd_max_pg_log_entries = 100000
osd_mon_heartbeat_interval = 30 #Performance tuning filestore
osd_mount_options_xfs =rw,noatime,logbsize=256k,delaylog
#osd_journal_size = 20480  日志大小，不指定，默认是5G
osd_op_log_threshold = 50
osd_min_pg_log_entries = 30000
osd_recovery_op_priority = 1
osd_mkfs_options_xfs = -f -i size=2048
osd_mkfs_type = xfs
osd_journal =/var/lib/ceph/osd/$cluster-$id/journal
journal_queue_max_ops = 500000
journal_max_write_bytes = 1048576000
journal_max_write_entries = 100000
journal_queue_max_bytes = 1048576000
objecter_infilght_op_bytes = 1048576000
objecter_inflight_ops = 819200
ms_dispatch_throttle_bytes = 1048576000
sd_data = /var/lib/ceph/osd/$cluster-$id
merge_threshold = 40
backfills = 1
mon_osd_min_down_reporters = 13
mon_osd_down_out_interval = 600
rbd_cache_max_dirty_object = 0
rbd_cache_target_dirty = 235544320
rbd_cache_writethrough_until_flush = false
rbd_cache_size = 335544320
rbd_cache_max_dirty = 335544320
rbd_cache_max_dirty_age = 60
rbd_cache = false　　5、安装ceph
　　#On All nodes to Install
yum -y install ceph　　admin节点上执行：
ceph-deploy mon create node1 node2 node3
ceph-deploy gatherkeys node1  #从moniter节点获得keys，用来管理节点　　

　　6、创建激活osd
ceph-deploy osd prepare node1:/osd0node1:/osd1 node1:/osd2 node2:/osd0 node2:/osd1 node2:/osd2 node3:/osd0node3:/osd1 node3:/osd2
ceph-deploy osd activate node1:/osd0node1:/osd1 node1:/osd2 node2:/osd0 node2:/osd1 node2:/osd2 node3:/osd0node3:/osd1 node3:/osd2
ceph-deploy admin node1 node2 node3   #从admin节点复制配置文件及key到node
chmod +r/etc/ceph/ceph.client.admin.keyring (所有节点为该文件添加读权限)　　也可以创建相关的pool
ceph osd pool create volumes 128
ceph osd pool create images 128
ceph osd pool create vms 128
ceph osd pool create backups 128

　　如果ceph安装失败，要清除ceph环境，执行以下命令
ceph-deploy purge node1 node2 node3 (所有节点)
ceph-deploy purgedata node1 node2 node3 (所有节点)
ceph-deploy forgetkeys