![Ceph - 纠删码池 - 总是获取不活跃的 pgs](https://linux22.com/image/735265/Ceph%20-%20%E7%BA%A0%E5%88%A0%E7%A0%81%E6%B1%A0%20-%20%E6%80%BB%E6%98%AF%E8%8E%B7%E5%8F%96%E4%B8%8D%E6%B4%BB%E8%B7%83%E7%9A%84%20pgs.png)
我正在尝试在 ceph 上实现与 raid6 类似的事情。
但是当我创建擦除编码池(k = 3 + m = 2(或k = 4))时,我总是得到不活动的pgs。
Ceph 健康详细信息如下:
HEALTH_WARN Reduced data availability: 128 pgs inactive
PG_AVAILABILITY Reduced data availability: 128 pgs inactive
pg 11.a is stuck inactive for 107.974251, current state unknown, last acting []
pg 11.b is stuck inactive for 107.974251, current state unknown, last acting []
pg 11.c is stuck inactive for 107.974251, current state unknown, last acting []
pg 11.d is stuck inactive for 107.974251, current state unknown, last acting []
pg 11.e is stuck inactive for 107.974251, current state unknown, last acting []
pg 11.f is stuck inactive for 107.974251, current state unknown, last acting []
pg 11.20 is stuck inactive for 107.974251, current state unknown, last acting []
EC 概况如下:
crush-device-class=
crush-failure-domain=rack
crush-root=default
jerasure-per-chunk-alignment=false
k=4
m=2
plugin=jerasure
technique=reed_sol_van
w=8
也尝试了 k=3,发生了同样的事情
粉碎规则(对于 k=4):
{
"rule_id": 1,
"rule_name": "r6pool",
"ruleset": 1,
"type": 3,
"min_size": 3,
"max_size": 6,
"steps": [
{
"op": "set_chooseleaf_tries",
"num": 5
},
{
"op": "set_choose_tries",
"num": 100
},
{
"op": "take",
"item": -1,
"item_name": "default"
},
{
"op": "chooseleaf_indep",
"num": 0,
"type": "rack"
},
{
"op": "emit"
}
]
}
因为如果不使用 EC 池,一切都很好,我很确定我不明白某些事情,但我无法指出我的错误。
为了在完全良好的 8 OSD 集群上获取此错误,我只需执行以下操作:
ceph osd erasure-code-profile set raid6 k=3 m=2
ceph osd pool create r6pool 128 128 erasure raid6
ceph -s 显示如下信息:
ceph -s
cluster:
id: 7635eaf1-df47-4bed-9cef-a3152cb4fa5f
health: HEALTH_WARN
Reduced data availability: 128 pgs inactive
services:
mon: 3 daemons, quorum CephMon1,CephMon3,CephMon2
mgr: CephClient(active)
osd: 9 osds: 8 up, 8 in
data:
pools: 2 pools, 256 pgs
objects: 0 objects, 0B
usage: 8.13GiB used, 31.8GiB / 40.0GiB avail
pgs: 50.000% pgs unknown
128 active+clean
128 unknown
我将非常感激您的帮助。谢谢。
答案1
嗯,一个:
crush-failure-domain=osd
解决了难题