On Mon, Jul 26, 2021 at 5:25 PM <cek+ceph@xxxxxxxxxxxx> wrote:
>
> Have found the problem. All this was caused by missing mon_host directive in ceph.conf. I have expected userspace to catch this, but it looks like it didn't care.
We should probably add an explicit check for that so that the error
message is explicit.
> We use DNS SRV in this cluster.
>
> With mon_host directive reinstated, it was able to connect:
> Jul 26 09:51:40 xx kernel: libceph: mon0 10.xx:6789 session established
> Jul 26 09:51:40 xx kernel: libceph: client188721 fsid 548a0823-815a-4ac5-a2e5-42cc7e8206ab
> Jul 26 09:51:40 xx kernel: rbd: image blk1: image uses unsupported features: 0x38
Now you just need to disable object-map, fast-diff and deep-flatten
with "rbd feature disable" as mentioned by Marc and Dimitri.
>
> I'm wondering what happens in case this mon1 host goes down, will the kernel module go through the remaining mon directive addresses?
Yes, these addreses are used to put together the initial list of
monitors (initial monmap). The kernel picks one at random and keeps
trying until the session with either of them gets established. After
that the real monmap received from the cluster replaces the initial
list.
Thanks,
Ilya
_______________________________________________
ceph-users mailing list -- ceph-users@xxxxxxx
To unsubscribe send an email to ceph-users-leave@xxxxxxx
