ceph 之 纠删码操作

通过命令行创建纠删码规则

• 首先，需要创建 erasure-code-profile ，当然，也可以使用默认的 erasure-code-profile ，列出现有的 erasure-code-profile ：
  
  # ceph osd erasure-code-profile ls default

  1 2	# ceph osd erasure-code-profile ls default

• 查看指定erasure-code-profile 的详细内容：
  
  # ceph osd erasure-code-profile get default k=2 m=1 plugin=jerasure technique=reed_sol_van

  1 2 3 4 5

  # ceph osd erasure-code-profile get default k=2 m=1 plugin=jerasure technique=reed_sol_van

• 自定义erasure-code-profile ， 创建一个只用hdd的 erasure-code-profile：
  
  # ceph osd erasure-code-profile set hdd-3-2 k=3 m=2 crush-device-class=hdd

  1	# ceph osd erasure-code-profile set hdd-3-2 k=3 m=2 crush-device-class=hdd

  
  可用的选项有：
  • crush-root: the name of the CRUSH node to place data under [default: default].
  • crush-failure-domain（故障域）: the CRUSH type to separate erasure-coded shards across [default: host].
  • crush-device-class（设备分类）: the device class to place data on [default: none, meaning all devices are used].
  • k and m (and, for the lrc plugin, l): these determine the number of erasure code shards, affecting the resulting CRUSH rule.
• 根据erasure-code-profile 创建crush rule：
  
  # ceph osd crush rule create-erasure erasure_hdd hdd-3-2 created rule erasure_hdd at 5

  1 2	# ceph osd crush rule create-erasure erasure_hdd hdd-3-2 created rule erasure_hdd at 5

• 查看crush rule：
  
  # ceph osd crush rule dump erasure_hdd { "rule_id": 5, "rule_name": "erasure_hdd", "ruleset": 5, "type": 3, "min_size": 3, "max_size": 5, "steps": [ { "op": "set_chooseleaf_tries", "num": 5 }, { "op": "set_choose_tries", "num": 100 }, { "op": "take", "item": -2, "item_name": "default~hdd" }, { "op": "chooseleaf_indep", "num": 0, "type": "host" }, { "op": "emit" } ] }

  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 26 27 28 29 30 31 32

  # ceph osd crush rule dump erasure_hdd { "rule_id": 5, "rule_name": "erasure_hdd", "ruleset": 5, "type": 3, "min_size": 3, "max_size": 5, "steps": [ { "op": "set_chooseleaf_tries", "num": 5 }, { "op": "set_choose_tries", "num": 100 }, { "op": "take", "item": -2, "item_name": "default~hdd" }, { "op": "chooseleaf_indep", "num": 0, "type": "host" }, { "op": "emit" } ] }

• 创建一个使用纠删码规则的pool
  
  # ceph osd pool create test-bigdata 256 256 erasure hdd-3-2 erasure_hdd pool 'test-bigdata' created

  1 2	# ceph osd pool create test-bigdata 256 256 erasure hdd-3-2 erasure_hdd pool 'test-bigdata' created

  
  语法： osd pool create <poolname> <int[0-]> {<int[0-]>} {replicated|erasure} [<erasure_code_profile>] {<rule>} {<int>}
  尽管crush rule 也是根据erasure_code_profile来创建的，但是这里创建纠删码pool的时候，还是需要明确指定erasure_code_profile的
  参考： http://docs.ceph.com/docs/master/rados/operations/pools/
• 调优：
  
  # ceph osd pool set test-bigdata fast_read 1 set pool 24 fast_read to 1

  1 2	# ceph osd pool set test-bigdata fast_read 1 set pool 24 fast_read to 1

  
  目前，这个fast_read 之针对纠删码池有效的
• 如果需要在该pool创建rbd，则需要：
  • 参考： http://docs.ceph.com/docs/master/rados/operations/erasure-code/
    
    # ceph osd pool set test-bigdata allow_ec_overwrites true set pool 24 allow_ec_overwrites to true

    1 2	# ceph osd pool set test-bigdata allow_ec_overwrites true set pool 24 allow_ec_overwrites to true

  • 创建一个replication pool来做cache tier
    
    # ceph osd pool create test-bigdata-cache-tier 128 pool 'test-bigdata-cache-tier' created # ceph osd tier add test-bigdata test-bigdata-cache-tier pool 'test-bigdata-cache-tier' is now (or already was) a tier of 'test-bigdata' # ceph osd tier cache-modetest-bigdata-cache-tier writeback set cache-mode for pool 'test-bigdata-cache-tier' to writeback # ceph osd tier set-overlay test-bigdata test-bigdata-cache-tier overlay for 'test-bigdata' is now (or already was) 'test-bigdata-cache-tier'

    1 2 3 4 5 6 7 8 9 10 11

    # ceph osd pool create test-bigdata-cache-tier 128 pool 'test-bigdata-cache-tier' created   # ceph osd tier add test-bigdata test-bigdata-cache-tier pool 'test-bigdata-cache-tier' is now (or already was) a tier of 'test-bigdata'   # ceph osd tier cache-modetest-bigdata-cache-tier writeback set cache-mode for pool 'test-bigdata-cache-tier' to writeback   # ceph osd tier set-overlay test-bigdata test-bigdata-cache-tier overlay for 'test-bigdata' is now (or already was) 'test-bigdata-cache-tier'

    
    其实，不仅纠删码池可以做cache tier，replication 池子也能做cache tier，因为，我们可能有一批ssd盘，我们就可以在ssd上创建pool来充当sas盘的cache tier以提高性能；结合纠删码、replication、sas、ssd，我们可以做出多种不同性能的存储用以应对不同的场景。
  • 然后 ceph 会提示： 1 cache pools are missing hit_sets  ， 还要设置 hit_set_count 和 hit_set_type
    
    # ceph osd pool set test-bigdata-cache-tier hit_set_count 1 set pool 29 hit_set_count to 1 # ceph osd pool set test-bigdata-cache-tier hit_set_type bloom set pool 29 hit_set_type to bloom

    1 2 3 4 5

    # ceph osd pool set test-bigdata-cache-tier hit_set_count 1 set pool 29 hit_set_count to 1   # ceph osd pool set test-bigdata-cache-tier hit_set_type bloom set pool 29 hit_set_type to bloom

通过编辑crushmap来添加规则

• 参考：https://phpor.net/blog/post/7080

参考：

• http://docs.ceph.com/docs/master/rados/operations/crush-map/#creating-a-rule-for-an-erasure-coded-pool
• http://blog.csdn.net/yangyimincn/article/details/50810229
• http://www.cnblogs.com/chris-cp/p/4791217.html
• 纠删码原理
• http://www.360doc.com/content/17/0814/11/46248428_679071890.shtml

实战中的问题：

1. 12个SAS在 60MB/s 的速度evict的时候，磁盘都很慢了，每个盘达到100左右的tps， 20MB/s左右的读写；比较坑的是，我基本没法控制evict的速度，只好静静地等待evict结束
2. evict 的同时还在promote， promote的速度倒是可控,但是 osd_tier_promote_max_bytes_sec 默认是5242880 字节（并不算很大）； 问题：池子已经没有写入了，为何还在evict和promote？
3. 修改cache-mode试试： 按说，修改cache-mode为proxy时，就不应该再出现evict和promote了
   
   果然，修改之后，ceph -s 立刻就看不到evict和promote了 🙂
4. 查看cache-mode:
   
   ceph osd pool ls detail|grep cache_mode

   1	ceph osd pool ls detail|grep cache_mode