ceph 指定OSD创建pool

https://my.oschina.net/wangzilong/blog/1549690

ceph集群中允许使用混合类型的磁盘，比如一部分磁盘是SSD，一部分是STAT。如果针对某些业务小高速磁盘SSD，某些业务需要STAT，在创建资源池的时候可以指定创建在某些OSD上。

基本步骤有8步：

当前只有STAT没有SSD，但是不影响实验结果。

1    获取crush map

[root@ceph-admin getcrushmap]# ceph osd getcrushmap -o /opt/getcrushmap/crushmap
got crush map from osdmap epoch 2482

2    反编译crush map

[root@ceph-admin getcrushmap]# crushtool -d crushmap -o decrushmap

3    修改crush map

在root default 后面添加下面两个bucket

root ssd {
	id -5
	alg straw
	hash 0
	item osd.0 weight 0.01
}
root stat {
        id -6
        alg straw
        hash 0
        item osd.1 weight 0.01
}

在rules部分添加如下规则：

rule ssd{
	ruleset 1
	type replicated
	min_size 1
	max_size 10
	step take ssd
	step chooseleaf firstn 0 type osd
	step emit
}
rule stat{
        ruleset 2
        type replicated
        min_size 1
        max_size 10
        step take stat
        step chooseleaf firstn 0 type osd
        step emit
}

4    编译crush map

[root@ceph-admin getcrushmap]# crushtool -c decrushmap -o newcrushmap

5    注入crush map

[root@ceph-admin getcrushmap]# ceph osd setcrushmap -i /opt/getcrushmap/newcrushmap
set crush map

[root@ceph-admin getcrushmap]# ceph osd tree
ID WEIGHT  TYPE NAME           UP/DOWN REWEIGHT PRIMARY-AFFINITY
-6 0.00999 root stat
 1 0.00999     osd.1                up  1.00000          1.00000
-5 0.00999 root ssd
 0 0.00999     osd.0                up  1.00000          1.00000
-1 0.58498 root default
-2 0.19499     host ceph-admin
 2 0.19499         osd.2            up  1.00000          1.00000
-3 0.19499     host ceph-node1
 0 0.19499         osd.0            up  1.00000          1.00000
-4 0.19499     host ceph-node2
 1 0.19499         osd.1            up  1.00000          1.00000
# 重新查看osd tree 的时候已经看见这个树已经变了。添加了名称为stat和SSD的两个bucket

6    创建资源池

[root@ceph-admin getcrushmap]# ceph osd pool create ssd_pool 8 8
pool 'ssd_pool' created
[root@ceph-admin getcrushmap]# ceph osd pool create stat_pool 8 8
pool 'stat_pool' created
[root@ceph-admin getcrushmap]# ceph osd dump|grep ssd
pool 28 'ssd_pool' replicated size 3 min_size 2 crush_ruleset 0 object_hash rjenkins pg_num 8 pgp_num 8 last_change 2484 flags hashpspool stripe_width 0
[root@ceph-admin getcrushmap]# ceph osd dump|grep stat
pool 29 'stat_pool' replicated size 3 min_size 2 crush_ruleset 0 object_hash rjenkins pg_num 8 pgp_num 8 last_change 2486 flags hashpspool stripe_width 0

注意：刚刚创建的两个资源池ssd_pool 和stat_pool 的 crush_ruleset  都是0，下面需要修改。

7    修改资源池存储规则

[root@ceph-admin getcrushmap]# ceph osd pool set ssd_pool crush_ruleset 1
set pool 28 crush_ruleset to 1
[root@ceph-admin getcrushmap]# ceph osd pool set stat_pool crush_ruleset 2
set pool 29 crush_ruleset to 2
[root@ceph-admin getcrushmap]# ceph osd dump|grep ssd
pool 28 'ssd_pool' replicated size 3 min_size 2 crush_ruleset 1 object_hash rjenkins pg_num 8 pgp_num 8 last_change 2488 flags hashpspool stripe_width 0
[root@ceph-admin getcrushmap]# ceph osd dump|grep stat
pool 29 'stat_pool' replicated size 3 min_size 2 crush_ruleset 2 object_hash rjenkins pg_num 8 pgp_num 8 last_change 2491 flags hashpspool stripe_width 0

# luminus 版本设置pool规则的语法是
[root@ceph-admin ceph]# ceph osd pool set ssd crush_rule ssd
set pool 2 crush_rule to ssd
[root@ceph-admin ceph]# ceph osd pool set stat crush_rule stat
set pool 1 crush_rule to stat

8    验证

验证前先看看ssd_pool 和stat_pool 里面是否有对象

[root@ceph-admin getcrushmap]# rados ls -p ssd_pool
[root@ceph-admin getcrushmap]# rados ls -p stat_pool
#这两个资源池中都没有对象

用rados命令 添加对象到两个资源池中

[root@ceph-admin getcrushmap]# rados -p ssd_pool put test_object1 /etc/hosts
[root@ceph-admin getcrushmap]# rados -p stat_pool put test_object2 /etc/hosts
[root@ceph-admin getcrushmap]# rados ls -p ssd_pool
test_object1
[root@ceph-admin getcrushmap]# rados ls -p stat_pool
test_object2
#对象添加成功

[root@ceph-admin getcrushmap]# ceph osd map ssd_pool test_object1
osdmap e2493 pool 'ssd_pool' (28) object 'test_object1' -> pg 28.d5066e42 (28.2) -> up ([0], p0) acting ([0,1,2], p0)
[root@ceph-admin getcrushmap]# ceph osd map stat_pool test_object2
osdmap e2493 pool 'stat_pool' (29) object 'test_object2' -> pg 29.c5cfe5e9 (29.1) -> up ([1], p1) acting ([1,0,2], p1)

上面验证结果可以看出，test_object1 存入osd.0中，test_object2 存入osd.1中。达到预期目的