Ceph Crush解析

转自:https://blog.csdn.net/u013525455/article/details/80795026

文章比较深入的写了CRUSH算法的原理和过程.通过调试深入的介绍了CRUSH计算的过程.

写在前面

读本文前,你需要对ceph的基本操作,pool和CRUSH map非常熟悉.并且较深入的读过源码.

分析的方法

首先,我们写了个c程序调用librados向pool中写入一个对象.然后使用 GDB(CGDB is recommended)来追踪CRUSH的运行过程.同时我们将会关注CRUSH相关的几个重要变量.以便于我们完全知道CRUSH源码是如何运行.

目录

如何追踪
编译CEPH 得到函数stack 追踪过程
input —> PGID PGID —> OSD set

1 如何追踪

1.1 编译CEPH

首先通过源码安装CEPH,然后使用./configure.如下添加编译参数/configure CFLAGS='-g3 –O0' CXXFLAGS='-g3 –O0'
-g3 意味着会产生大量的调试信息.-O0 非常重要, 它意味着关闭编译器的优化,如果没有,使用GDB追踪程序时,大多数变量被优化,无法显示。配置后,make and sudo make install。附: -O0只适合用于实验情况,在生产环境中编译器优化是必须进行的。

1.2 得到函数stack

众所周知,CRSUSH的核心函数是 crush_do_rule(位置 crush/mapper.c line 779).

?

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

<code class="language-c++ hljs java">

/**

 * crush_do_rule - calculate a mapping with the given input and rule

 * @map: the crush_map

 * @ruleno: the rule id

 * @x: hash input

 * @result: pointer to result vector

 * @result_max: maximum result size

 * @weight: weight vector (for map leaves)

 * @weight_max: size of weight vector

 * @scratch: scratch vector for private use; must be >= 3 * result_max

 */

int crush_do_rule(const struct crush_map *map,

          int ruleno, int x, int *result, int result_max,

          const __u32 *weight, int weight_max,

          int *scratch)

</code>

通过这个函数将crush计算过程分为两部分:
1. input -> PGID
2. PGID -> OSD set.
第一部分,使用GDB来得到函数过程

通过 -g 参数来编译例子程序 gdb来参看 rados_write 然后,在添加断点b crush_do_rule前,进入GDB的接口. 在函数crush_do_rule 处停留 得到函数stack,然后使用GDB log将调试信息输出到文件中 .
下面让我们来深入的研究这个过程.

函数stack和下面的相似

?

1

2

3

4

5

6

7

8

9

10

11

12

13

<code class="hljs vala">#12 main

#11 rados_write

#10 librados::IoCtxImpl::write

#9 librados::IoCtxImpl::operate

#8 Objecter::op_submit

#7 Objecter::_op_submit_with_budget

#6 Objecter::_op_submit

#5 Objecter::_calc_target

#4 OSDMap::pg_to_up_acting_osds

#3 OSDMap::_pg_to_up_acting_osds

#2 OSDMap::_pg_to_osds

#1 CrushWrapper::do_rule

#0 crush_do_rule</code>

追踪过程

CRUSH 计算过程总结如下:

?

1

<code class="hljs applescript"> INPUT(object name & pool name) —> PGID —> OSD set.</code>

本文中主要关注 计算过程

2.1 input —> PGID

你可以按顺序阅读源码.最后,通过列出转换关键过程
从input 到 pgid

?

1

2

<code class="language-c++ hljs objectivec">Ceph_hash.h (include):

extern unsigned ceph_str_hash(int type, const char *s, unsigned len);</code>

?

1

<code class="hljs avrasm">OSDMap.h (osd):    int ret = object_locator_to_pg(oid, loc, pg);</code>

首先从 rados_ioctx_create, 和lookup_pool 通过pool名字得到 poolid. 把poolid封装进librados::IoCtxImpl 类型变量ctx;

然后在rados_write 对象名被封装进oid; 然后在librados::IoCtxImpl::operateoid 和oloc(comprising poolid) 被包装成Objecter::Op * 类型变量objecter_op;

通过各种类型的封装我们到到 _calc_target 这层. 我们得到不断的oid 和poolid. 然后读取目标 pool 的信息.

这里写图片描述

这里写图片描述

(in my cluster, pool “neo” id is 29, name of object to write is “neo-obj”)<喎�"/kf/ware/vc/" target="_blank" class="keylink">vcD4NCjxwPtTaPGNvZGU+b2JqZWN0X2xvY2F0b3JfdG9fcGc8L2NvZGU+LCC12tK7tM68xsvjtNM8Y29kZT5jZXBoX3N0cl9oYXNoPC9jb2RlPiC5/s+jttTP88P719azyc6q0ru49nVpbnQzMl90IMDg0M2x5MG/LNKyvs3Kx8v5zr21xCA8Y29kZT5wczwvY29kZT4gPHN0cm9uZz4ocGxhY2VtZW50IHNlZWQpPC9zdHJvbmc+PC9wPg0KPHByZSBjbGFzcz0="brush:java;">unsigned int ceph_str_hash(int type, const char *s, unsigned int len) { switch (type) { case CEPH_STR_HASH_LINUX: return ceph_str_hash_linux(s, len); case CEPH_STR_HASH_RJENKINS: return ceph_str_hash_rjenkins(s, len); default: return -1; } }

这里写图片描述

然后得到 PGID. 以前我认为pgid 是单一变量,然而不是. PGID 是个包含 poolid 和 ps 的结构体变量.

这里写图片描述

?

1

2

3

4

5

6

7

8

9

10

11

12

<code class="language-c++ hljs r">//pgid  不仅仅是一个数字,还有好多信息

// placement group id

struct pg_t

{

  uint64_t m_pool;

  uint32_t m_seed;

  int32_t m_preferred;

  pg_t() : m_pool(0), m_seed(0), m_preferred(-1) {}

  pg_t(ps_t seed, uint64_t pool, int pref=-1) :

  ...

还有很多信息略</code>

crush_do_rule 的输入参数x是什么?让我们继续风雨兼程. 然后在 _pg_to_osds 有一行 ps_t pps = pool.raw_pg_to_pps(pg); //placement ps. pps 就是 x.
PPS 如何计算? 在函数crush_hash32_2(CRUSH_HASH_RJENKINS1,ceph_stable_mod(pg.ps(), pgp_num, pgp_num_mask),pg.pool());

?

1

2

3

4

5

6

7

8

9

<code class="hljs go">__u32 crush_hash32_2(int type, __u32 a, __u32 b)

{

    switch (type) {

    case CRUSH_HASH_RJENKINS1:

        return crush_hash32_rjenkins1_2(a, b);

    default:

        return 0;

    }

}</code>

这里写图片描述

ps mod pgp_num_mask 的结果(例如 a) 和poolid(例如 b) 进行哈希. 这就是pps,也就是x

这里写图片描述

我们得到第二个过程输入参数的 x.第一阶段过程图如下图

第一阶段过程图

这里写图片描述
P.S. you can find something in PG’s name and object name.

2.2 PGID —> OSD set

首先需要明白几个概念

weight VS reweight
这里写图片描述
这里,“ceph osd crush reweight” 设置了OSD的权重

weight
这个重量为任意值(通常是磁盘的TB大小,1TB设置为1),并且控制系统尝试分配到OSD的数据量。

reweight
reweight将覆盖了weight量。这个值在0到1的范围,并强制CRUSH重新分配数据。它不改变buckets 的权重,并且是CRUSH不正常的情况下的纠正措施。(例如,如果你的OSD中的一个是在90%以上,其余为50%,可以减少权重,进行补偿。)

primary-affinity

主亲和力默认为1(即,一个OSD可以作为主OSD)。primary-affinity变化范围从0到1.其中0意味着在OSD可能用作主OSD,设置为1则可以被用作主OSD.当其<1时,CRUSH将不太可能选择该OSD作为主守护进程。

pg 和 pgp的区别
- PG = Placement Group
- PGP = Placement Group for Placement purpose
- pg_num = number of placement groups mapped to an OSD

当增加每个pool的pg_num数量时,每个PG分裂成半,但他们都保持到它们的父OSD的映射。

直到这个时候,ceph不会启动平衡策略。现在,当你增加同一池中的pgp_num值,PGs启动从父OSD迁移到其他OSD,ceph开始启动平衡策略。这是PGP如何起着重要的作用的过程。

pgp-num是在CRUSH算法中使用的参数,不是 pg-num.例如pg-num = 1024 , pgp-num = 1.所有的1024个PGs都映射到同一个OSD.当您增加PG-NUM是分裂的PG,如果增加PGP-NUM将移动PGs,即改变OSD的map。PG和PGP是很重要的概念

?

1

2

<code class="hljs cpp">void do_rule(int rule, int x, vector<int>& out, int maxout,

           const vector<__u32>& weight) const {}</int></code>

在了解了这些概念后开始第二部分
PGID -> OSD set. 现在我们在 do_rule: void do_rule(int rule, int x, vector& out, int maxout, const vector<__u32>& weight)

do_rule源代码

?

1

2

3

4

5

6

7

8

9

10

<code class="hljs cpp">  void do_rule(int rule, int x, vector<int>& out, int maxout,

           const vector<__u32>& weight) const {

    Mutex::Locker l(mapper_lock);

    int rawout[maxout];

    int scratch[maxout * 3];

    int numrep = crush_do_rule(crush, rule, x, rawout, maxout, &weight[0], weight.size(), scratch);

    if (numrep < 0)

      numrep = 0;

    out.resize(numrep);

    for (int i=0; i<numrep; code=""></numrep;></int></code>

让我们看下输入参数
x 就是我们已经得到的 pps rule 就是内存中的crushrule’s number(不是 ruleid, 在我的crushrule set中, this rule’s id是 3), weight 是已经讲过的 reweight 变化范围从1 到 65536. 我们定义了 rawout[maxout] 来存储 OSD set, scratch[maxout * 3] 为计算使用. 然后我们进入了crush_do_rule.

这里写图片描述

这里写图片描述

PGID -> OSDset OUTLINE

下面要仔细研究3个函数, firstn 意味着副本存储, CRUSH 需要去选择n
个osds存储副本. indep是纠删码存储过程.我们只关注副本存储方法

crush_do_rule: 反复do crushrules crush_choose_firstn: 递归选择特定类型的桶或设备 crush_bucket_choose: 直接选择bucket的子节点

crush_do_rule

首先这是我的crushrule 和集群的层次结构
这里写图片描述
参数

?

1

2

3

4

5

6

7

8

9

10

11

12

<code class="hljs cpp"><code class="hljs perl">@map: the crush_map

@ruleno: the rule id

@x: hash input

@result: pointer to result vector

@result_max: maximum result size

@weight: weight vector (for map leaves)

@weight_max: size of weight vector

@scratch: scratch vector for private use; must be >= 3 * result_max

int crush_do_rule(const struct crush_map *map,

          int ruleno, int x, int *result, int result_max,

          const __u32 *weight, int weight_max,

          int *scratch)</code></code>

值得说的变量emit 通常用在规则的结束,同时可以被用在在形相同规则下选择不同的树。

 

scratch[3 * result_max]

?

1

2

3

<code class="hljs cpp"><code class="hljs perl">int *a = scratch;

int *b = scratch + result_max;

int *c = scratch + result_max*2;</code></code>

a, b, c 分别指向 scratch向量的0, 1/3, 2/3的位置.
w = a; o = b;
w被用作一个先入先出队列来在CRUSH map中进行横向优先搜索(BFS traversal).
o存储crush_choose_firstn选择的结果.
c存储最终的OSD选择结果.
crush_choose_firstn计算后如果结果不是OSD类型, o 交给w.以便于 w成为下次crush_choose_firstn的输入参数. 如上所述, crush_do_rule 反复进行 crushrules 迭代. 你可以在内存中发现规则:

这里写图片描述

过程步骤

step 1 put root rgw1 in w(enqueue);

step 2 would run crush_choose_firstn to choose 1 rack-type bucket from root rgw1

下面分析crush_choose_firstn过程

crush_choose_firstn 函数

这个函数递归的选择特定bucket或者设备,并且可以处理冲突,失败的情况.
如果当前是choose过程,通过调用crush_bucket_choose来直接选择.
如果当前是chooseleaf选择叶子节点的过程,该函数将递归直到得到叶子节点.

crush_bucket_choose 函数

crush_bucket_choose是CRUSH最重要的函数.应为默认的bucket类型是straw,常见的情况下我们会使用straw类型bucket,然后就会进入bucket_straw_choose

case进行跳转

?

1

2

<code class="hljs cpp"><code class="hljs cs">case CRUSH_BUCKET_STRAW:   

return bucket_straw_choose((struct crush_bucket_straw *)in,</code></code>

完整代码

?

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

<code class="hljs cpp"><code class="language-c++ hljs cs">static int crush_bucket_choose(struct crush_bucket *in, int x, int r)

{

    dprintk(" crush_bucket_choose %d x=%d r=%d\n", in->id, x, r);

    BUG_ON(in->size == 0);

    switch (in->alg) {

    case CRUSH_BUCKET_UNIFORM:

        return bucket_uniform_choose((struct crush_bucket_uniform *)in,

                      x, r);

    case CRUSH_BUCKET_LIST:

        return bucket_list_choose((struct crush_bucket_list *)in,

                      x, r);

    case CRUSH_BUCKET_TREE:

        return bucket_tree_choose((struct crush_bucket_tree *)in,

                      x, r);

    case CRUSH_BUCKET_STRAW:

        return bucket_straw_choose((struct crush_bucket_straw *)in,

                       x, r);

    case CRUSH_BUCKET_STRAW2:

        return bucket_straw2_choose((struct crush_bucket_straw2 *)in,

                        x, r);

    default:

        dprintk("unknown bucket %d alg %d\n", in->id, in->alg);

        return in->items[0];

    }

}</code></code>

?

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

<code class="hljs cpp"><code class="language-c++ hljs cs">/* straw */

static int bucket_straw_choose(struct crush_bucket_straw *bucket,

                   int x, int r)

{

    __u32 i;

    int high = 0;

    __u64 high_draw = 0;

    __u64 draw;

    for (i = 0; i < bucket->h.size; i++) {

        draw = crush_hash32_3(bucket->h.hash, x, bucket->h.items[i], r);

        draw &= 0xffff;

        draw *= bucket->straws[i];

        if (i == 0 || draw > high_draw) {

            high = i;

            high_draw = draw;

        }

    }

    return bucket->h.items[high];

}

</code></code>

bucket结构体

?

1

2

3

4

5

6

<code class="hljs cpp"><code class="language-c++ hljs cs">struct crush_bucket_straw {

    struct crush_bucket h;

    __u32 *item_weights;   /* 16-bit fixed point */

    __u32 *straws;         /* 16-bit fixed point */

};

</code></code>

这里写图片描述

可以看到 bucket root rgw1’s id是 -1, type = 10 意味着根节点 alg = 4 意味着 straw 类型. 这里 weight 是 OSD权重 we set scales up by 65536(i.e. 37 * 65536 = 2424832). 然后看下循环 T: 对每个输入的son bucket , 例如升上图rack1crush_hash32_3 hashes x,bucket id(rack1’s id)r(current selection’s order number), 这3个变量是a uint32_t 类型变量, 结果 & 0xffff, 然后乘straw(rack1’s straw value, straw calculation seen below), 最后得到这个值, 在一次循环中, for one son bucket(rack1 here). 我们在循环中计算每个 son bucket然后选择最大的 . 然后一个son bucket 被选择 . Nice job! 下面是个计算的例子

这里写图片描述

调用层次,图表描述

这里写图片描述

这里写图片描述

这里写图片描述

这里写图片描述

这里写图片描述

结论

我们已经研究了CRUSH计算中的重要部分,其余的部分就是迭代和递归,直到选择了所有的OSD.

关于 straw 值

详细的代码在 src/crush/builder.c crush_calc_straw.
总之,straw 值总是和OSD权重正相关.straw2正在开发.

猜你喜欢

转载自blog.csdn.net/majianting/article/details/86638935