|
|
[root@ceph2 ceph]# ceph -s1 A Q6 S4 Q; O0 n$ u
cluster:
% e, y9 C$ k8 `% Q6 G; @1 ] id: 2f6bb77f-8b7e-4530-bd4b-8e6f6ab58833
3 t y: g9 ~* g+ p8 W0 q: P( Z health: HEALTH_WARN4 }+ P1 Q* m4 X% R" W( M3 [4 i- {
all OSDs are running pacific or later but require_osd_release < pacific4 g2 S" m; i8 F" o
. v: @# u" Y( m! O& } services:8 k' F. V$ j! Z# s7 l
mon: 3 daemons, quorum ceph1,ceph2,ceph3 (age 42m)( q _2 s0 [. G" m" q6 W+ o- b1 C/ K
mgr: ceph3(active, since 12m), standbys: ceph2, ceph1. [. T9 ~- p/ t. R/ J
mds: 1/1 daemons up, 2 standby2 \' J+ h$ i; n7 a, s% R+ @! _0 Z2 C
osd: 9 osds: 9 up (since 36m), 9 in (since 10h)- k6 H; T2 K' R0 x% L5 G5 ?5 s
rgw: 3 daemons active (3 hosts, 1 zones)
& m* J8 o4 y7 A% j6 ~ U4 P3 \" s4 Z& N. v3 i
data:
; H4 g4 r% p6 Z' _% i' K volumes: 1/1 healthy
/ W Q/ M6 { S9 u+ i2 Y pools: 9 pools, 289 pgs' a$ I' A- ?1 Y2 I' S1 j, I: S2 F/ W) b
objects: 735 objects, 2.0 GiB
! M* r. k* w% [: O2 k6 i usage: 51 GiB used, 174 GiB / 225 GiB avail1 p6 T" L: _1 i3 d' O* X, X" S
pgs: 289 active+clean
' M! G' A% R% e9 p0 c$ Z1 V: H1 S8 _+ n+ Y$ B3 }% d8 n
直译所有 OSD 实际运行的版本已经是 Pacific 或更高( w6 t3 M! o# q8 ]
但集群配置里仍然要求 require_osd_release < pacific
人话OSD 已经升级完了，但集群“版本锁”还停留在老版本
Ceph 为了防止误升级导致数据不可逆，升级是两阶段的：
5 S+ }; W9 T+ H/ q先升级 daemon（你已经完成） 再手动提升集群最小版本要求（你还没做）
9 k0 y8 J' t, u3 n8 O
: @9 L1 Y* h- W g4 r# z5 k$ ? l3 W6 H8 M: O
[root@ceph2 ceph]# ceph osd require-osd-release pacific
, G& {- R0 H: J, z/ T% \( x3 D2 W5 C& i
再次查看状态：
! j$ t& ]) r3 H# i; }[root@ceph2 ceph]# ceph -s) _0 r, Z) d7 ^4 X9 x
cluster:( U& }' l/ [- k* r
id: 2f6bb77f-8b7e-4530-bd4b-8e6f6ab588336 n9 o3 z- E% V, n V% k" P
health: HEALTH_OK; @5 w' V" r- G# w5 M, C
- `4 ~% [3 y6 F, f
services:
) |0 B( D5 j5 n; b$ O mon: 3 daemons, quorum ceph1,ceph2,ceph3 (age 43m)& `4 u) B- U/ e- {# x+ e
mgr: ceph3(active, since 13m), standbys: ceph2, ceph1
$ @2 ^& e# V4 q& i$ a! S mds: 1/1 daemons up, 2 standby4 Y$ b6 d7 i+ Y( K: n
osd: 9 osds: 9 up (since 37m), 9 in (since 10h)3 S/ D1 _. H8 z. @. V: J
rgw: 3 daemons active (3 hosts, 1 zones)0 [! ]3 f! s& [' U: ~0 I$ G
2 a) X; c* k, g2 t9 P data:
( ]6 ^: ^) e1 D volumes: 1/1 healthy. `+ X- |2 o- B) M+ K
pools: 9 pools, 289 pgs+ N8 }$ u' @) [0 \$ x2 H% T
objects: 735 objects, 2.0 GiB+ Q8 }: Y g: T! x
usage: 51 GiB used, 174 GiB / 225 GiB avail
( ^% k3 w" C# ?$ C8 O' T pgs: 289 active+clean% T1 K' W* P2 h9 j( I
3 y& i- e( q8 \1 B
: f! R( h) G( ]+ s 这是一次性、不可回退操作
& }7 g1 S) V3 U" q; O但你的前提已经满足（全部 OSD 都是 pacific+），所以是安全的。9 g. i: @, n' `0 o5 ~( X
7 }5 X& M; n$ O+ p
% I- U6 h* r w8 u' @: W2 b# ceph osd versions% p+ }$ _& s. z1 ~& w
{4 ^' E5 \$ c- `5 e# p$ Q- z- j
"ceph version 16.2.10-275.el8cp (69b25b314048bb8542dce8512a249fc34a7add1d) pacific (stable)": 9# c: ]! q' y: ^& |0 r0 q# m
}% \- `$ D7 |( z5 j" T
2 R7 ^: H* T2 j+ J2 n1 j
8 L* b/ s7 ~. z
- W. k( ~* v6 ]; \
+ ^) _8 u0 Y, ]8 ]/ W
X& E E' [2 q6 j! v |
|
发表于 2025-12-30 00:15:56