|
|
[root@ceph2 ceph]# ceph -s
% r6 T1 k& Z/ j' `1 C' v cluster:5 H. @/ q; A1 p
id: 2f6bb77f-8b7e-4530-bd4b-8e6f6ab58833 K# {5 V! q( Q- F# d, H
health: HEALTH_WARN
- \5 e& }( K8 ^: c! N6 k z3 A( } all OSDs are running pacific or later but require_osd_release < pacific
+ [% K. [$ ~) }) R2 o# X7 F( I
. w; G6 \1 Q5 ]9 I" O3 K* E4 B services:
0 w$ t X' j2 h. z mon: 3 daemons, quorum ceph1,ceph2,ceph3 (age 42m)
! `+ w/ y2 {4 z! d! ] mgr: ceph3(active, since 12m), standbys: ceph2, ceph1; m) ~& J# ~5 h2 D
mds: 1/1 daemons up, 2 standby) [, r6 ?1 k" v, N. j J
osd: 9 osds: 9 up (since 36m), 9 in (since 10h)
- W( }$ A; q; s rgw: 3 daemons active (3 hosts, 1 zones); G7 L1 t- y% j4 @ g/ m: l/ `
6 h! o! U/ p& G6 V6 F }; S0 p. A* C data:
; Y0 ~1 i5 x" [- [* `8 @: A volumes: 1/1 healthy
9 c' z# @' Z, T& A# ^$ p' x pools: 9 pools, 289 pgs# P3 [6 k4 w3 x# ~' w
objects: 735 objects, 2.0 GiB5 b8 i* O) i ]3 O, p$ ?
usage: 51 GiB used, 174 GiB / 225 GiB avail$ B7 K0 U9 S/ @2 ~; H6 Y8 o6 ^
pgs: 289 active+clean p5 G* y: `9 W p' F6 R
5 e/ a! C, C3 {5 A. t+ \
直译所有 OSD 实际运行的版本已经是 Pacific 或更高& W7 u+ ?# g1 h- Y- }7 R: F7 A
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： : j- ^1 I2 w, ?
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） $ J6 T9 k. O8 r/ k
$ ?: @/ r7 S! O# @" U; \! c0 V
# n T9 d' @8 u4 q5 I! M' x[root@ceph2 ceph]# ceph osd require-osd-release pacific6 P& w# }7 ?, s q$ y+ Q/ j$ O, y
( g1 U. B5 G: v1 _; ~" p! w* e
再次查看状态：% l8 r# E) w3 `" O$ x1 q
[root@ceph2 ceph]# ceph -s
7 l) Z; o; z, c5 ?7 H cluster:
6 V. z$ h1 [* [+ I% k* u id: 2f6bb77f-8b7e-4530-bd4b-8e6f6ab58833
1 v, d% D) r4 P/ @, P; i2 \ health: HEALTH_OK( [- w' q) L1 F0 [6 G# H3 _- ^
& w+ [1 i# _, p+ s0 N services:" [; f% e0 _; j6 N& U1 H
mon: 3 daemons, quorum ceph1,ceph2,ceph3 (age 43m)
9 n# p$ _; l7 @2 O6 L! E mgr: ceph3(active, since 13m), standbys: ceph2, ceph1
" q3 e- F+ \, r0 ~ mds: 1/1 daemons up, 2 standby
/ c: m+ w1 }% V; A" G7 s osd: 9 osds: 9 up (since 37m), 9 in (since 10h)
4 N; d1 o5 C: E) w j0 ` rgw: 3 daemons active (3 hosts, 1 zones)
t' L1 Q5 i$ f; G a) W9 M$ r2 [4 K) a# \- I0 R1 M1 j. l
data:3 x% P) D) \9 n/ I& O
volumes: 1/1 healthy
# r( B, T5 L$ r8 U3 A! w2 t pools: 9 pools, 289 pgs5 h, U( m# I/ S9 J& M0 J( X" e
objects: 735 objects, 2.0 GiB, W, w# b N, { D
usage: 51 GiB used, 174 GiB / 225 GiB avail6 ^: t3 D- {7 |, G3 q7 z
pgs: 289 active+clean; o- H! J5 u, [
3 C, m( {; g0 M. X0 r6 s# G) d
: G# Q. S1 Q6 w5 g
这是一次性、不可回退操作
' D; Q( x& {; _ M$ f8 P( v1 H但你的前提已经满足（全部 OSD 都是 pacific+），所以是安全的。
; k- F, r- G% g' A1 v' L! ]2 I$ f8 ?* s6 n& X, e" d2 i& T4 T8 \. M j
" k6 X# V: {5 ]% Z' N# ceph osd versions$ r1 e5 m8 K: v e
{7 Z8 q7 G) P; H( r! g3 Y$ K
"ceph version 16.2.10-275.el8cp (69b25b314048bb8542dce8512a249fc34a7add1d) pacific (stable)": 9# M5 i9 X/ T0 Y4 t S% J
}) o4 @' l! X1 ?. u$ x5 ?
, k1 `3 p: D2 S+ s: N8 U2 e) L! R# K! |5 p: c, B$ b
- R U/ ^; O/ _: h z" t: k
% q5 T' A- e+ B/ [2 W: V
2 L4 Q# W; R4 f+ g |
|
发表于 2025-12-30 00:15:56