如何删除ceph集群中无效的osd？

Question

[root@dev-master ceph-cluster]# ceph osd tree
ID WEIGHT  TYPE NAME     UP/DOWN REWEIGHT PRIMARY-AFFINITY
-1 0.01740 root default
-4 0.00580     host osd2
 0 0.00580         osd.0    down        0          1.00000
-5 0.00580     host osd3
 1 0.00580         osd.1    down        0          1.00000
-6 0.00580     host osd1
 2 0.00580         osd.2    down        0          1.00000
 5       0 osd.5              up        0          1.00000
[root@dev-master ceph-cluster]# ceph osd out 5
osd.5 is already out.
[root@dev-master ceph-cluster]# ceph osd crush remove osd.5
device 'osd.5' does not appear in the crush map
[root@dev-master ceph-cluster]# ceph auth del osd.5
entity osd.5 does not exist
[root@dev-master ceph-cluster]# ceph osd rm 5
Error EBUSY: osd.5 is still up; must be down before removal.

但我在任何主机上都找不到osd.5 5。

Answer 1

您可以尝试手动标记osd；如果osd进程在某个地方实际运行，它将在几秒钟后重新标记自己。

ceph osd down osd.5; ceph osd rm "$_"

Answer 2

有时ceph osd purge [osd.daemon]可以工作。

Answer 3

毕竟，它是给定节点上的一个服务，所以我建议先停止这个服务。

systemctl disable ceph-osd@5.service

那就跟著

ceph osd out osd.5
ceph osd safe-to-destroy osd.5
ceph osd destroy osd.5 --yes-i-really-mean-it
ceph osd crush remove osd.5
ceph osd rm osd.5