|
|
尝试二、修复down掉的osd
$ ~: I/ k1 ?+ M/ d* T- `% ~8 T2 S6 Y! \& }" J4 @' e2 ]
该方法主要应用于某个osd物理损坏，导致激活不了! H0 { u. N: ^5 s
! Z! O8 q9 c7 `& X' A
1、查看osd树
& b- U) Q1 ?' d. t, D1 @/ N$ D6 a6 E
复制代码. J: `/ y/ Z) G' F# }4 _' }
root@ceph01:~# ceph osd tree( E2 c# A; Z: t
ID WEIGHT TYPE NAME UP/DOWN REWEIGHT PRIMARY-AFFINITY # n) r& c0 ~& i* j
-1 0.29279 root default 8 F+ }$ F# h$ M0 T8 m
-2 0.14639 host ceph01 * d1 k, |6 C6 R& U6 T8 Q) I
0 0.14639 osd.0 up 1.00000 1.00000
# d- n" ?% B+ I( V O" X" J# ]-3 0.14639 host ceph02
0 f& U+ a% D% b$ i) x. | 1 0.14639 osd.1 down 0 1.00000
( v$ V3 }; R% a4 Q$ E& h复制代码0 P- q6 C/ p! u E; f" j X# B' S
发现osd.1是down掉的。% ~+ R* @8 |4 ~' w
5 [' S4 G$ K! r* \: T. W% w& ?6 M' _2、将osd.1的状态设置为out
f& e H; K: |( q5 n) j
- k/ W. n: t$ r, i+ ~, rroot@ceph02:~# ceph osd out osd.1
4 s: D+ a3 E% y! j+ Zosd.1 is already out.
9 ^) E- S O9 n( {1 C# t" X6 G3、从集群中删除
J* d3 b: U* I
" w9 `, |$ Q; _) t: S+ |root@ceph02:~# ceph osd rm osd.1
" E/ h9 \# ?1 }7 ~# ]4 a; c9 \$ sremoved osd.1
( I9 p6 D" Y) G- t/ `( v4、从CRUSH中删除
5 C& |. g( K. q8 I1 [3 D E g. B
root@ceph02:~# ceph osd crush rm osd.1 3 K4 \% W1 i7 Y
removed item id 1 name 'osd.1' from crush map5 d1 `& ^8 C) z+ v T
5、删除osd.1的认证信息
) P. T6 e |, Y
# [& I3 ~: z `. @" hroot@ceph02:~# ceph auth del osd.1
) B7 d) w3 U- W* @+ O( K. h( L9 Uupdated
9 M+ H+ V$ A- v' J9 g. p6、umount
( [2 l- x0 W% c
& s+ r' _* l L3 B- I7 wumount /dev/sdb1
) M j1 g7 M0 A! a) L0 I! E . P0 @* \7 @" O- J4 G7 W' R- a
2 M% `. Y/ n2 c; u+ `" h7、再次查看osd的集群状态
" }5 L q$ R$ Z1 V. W; P8 j7 e1 o% C5 F! }0 I
复制代码/ l) k& _, l8 N( u
root@ceph02:~# ceph osd tree! D1 n! K; I7 h8 `: S
ID WEIGHT TYPE NAME UP/DOWN REWEIGHT PRIMARY-AFFINITY ( r. |* P" a8 d8 P2 J
-1 0.14639 root default + S( I- O1 d; e' g5 r+ ]- W! z
-2 0.14639 host ceph01
4 d, z$ m. v& }8 H 0 0.14639 osd.0 up 1.00000 1.00000
8 V+ A% n* m2 B-3 0 host ceph02
! h% e3 `( l- g) j复制代码" T9 v4 V5 e2 b8 T
8、登录ceph-deploy节点
, k2 [: l6 R2 Y: @) N. s0 Z. a! o5 f, P% h
root@ceph01:~# cd /root/my-cluster/5 @- v6 \: R' S
root@ceph01:~/my-cluster# ! _. z3 Z* F( m; Z
9、初始化磁盘
# v0 G/ c# ]% O$ n, b; Q% U' K8 Z# R6 C* M: `
4 Z8 p9 s3 V$ r+ v6 j
ceph-deploy --overwrite-conf osd prepare ceph02:/dev/sdb1
' P- g& j3 y' Z) ^% u3 s0 a+ c
+ N) ^* p" t8 u- A在后面的版本中需要重新执行添加osdceph-deploy osd create node1 --data /dev/sdb
) M+ K8 H- `+ N& H% ?% y7 X# g6 m# G o% t
10、再次激活所有的osd(记住是所有的，不只是down掉这一个)
2 W: l- z* k& E7 C& M3 ~' T/ T8 v1 g5 ~
ceph-deploy osd activate ceph01:/dev/sdb1 ceph02:/dev/sdb1. F; P* t+ q4 u% ^4 z) F9 U4 M1 z
11、查看osd树和健康状态2 p) x8 `/ R9 ]. K9 }
2 d/ D6 a- o6 t: E6 [6 M
复制代码8 i0 G3 b& P5 m5 O( M
root@ceph01:~/my-cluster# ceph osd tree1 B# ~' R9 p5 e5 J
ID WEIGHT TYPE NAME UP/DOWN REWEIGHT PRIMARY-AFFINITY * q! {, k* }" Q# e" k
-1 0.29279 root default
# W5 Z4 H, J( J- X; m-2 0.14639 host ceph01 : D: m5 R2 q1 P8 r5 ?
0 0.14639 osd.0 up 1.00000 1.00000
6 H; W: R0 c! r% d7 {7 P-3 0.14639 host ceph02
' |5 C7 b. ~! h% W% w5 D+ j 1 0.14639 osd.1 up 1.00000 1.00000 - }& g* w& O0 O" Y9 t8 l' _) e
root@ceph01:~/my-cluster#
. U8 Q0 q _! D6 Z0 O# g' }: D复制代码# W m. U/ x; L: N' v
复制代码/ k4 O* \9 `2 y* T5 g" R# t* |! r t" _. A
root@ceph01:~/my-cluster# ceph -s
" i: X1 M# ~# \3 t' w3 \6 h1 }. b cluster ecacda71-af9f-46f9-a2a3-a35c9e51db9e0 R0 s$ Q0 Z e! Q* A
health HEALTH_OK
5 J/ D$ Q( ^7 v) o3 f& a- j+ f4 O monmap e1: 1 mons at {ceph01=10.111.131.125:6789/0}
. C; d+ g% Q. y7 G* n' L% t! V. n election epoch 14, quorum 0 ceph01
; t& b7 U( z& [5 {" _8 ^, E osdmap e150: 2 osds: 2 up, 2 in
) ]# H% r1 Q! V$ `0 v6 \( U flags sortbitwise,require_jewel_osds
; c% r" T! }2 [5 m pgmap v9284: 64 pgs, 1 pools, 17 bytes data, 3 objects! _0 `( _9 G- W# ?- n5 D/ C2 ?
10310 MB used, 289 GB / 299 GB avail0 S6 ^ A0 n8 P
64 active+clean* f- q/ I- V4 D6 S% q1 y/ k3 `! L$ B& Y
复制代码
4 H4 b9 ? S$ X, f$ f- e- z只有为 HEALTH_OK 才算是正常的。
& K. Q) q8 k# B$ ]
7 d$ ~$ D- w0 ]1 R B3 Q0 T
4 e3 r N( K. d) s" [ |
|
发表于 2018-10-6 15:47:19