lxc ceph-fuse

[Kiotsoukis]

Hi,

I'm trying to mount a ceph fs directory in a lxc container. But I don't get past

root@lxc-kio:~# ceph-fuse -m xx.xx.xx.xx --id kio -r /kio /mntkio
ceph-fuse[319]: starting ceph client
2021-08-27 10:54:58.787349 7f6897ee4100 -1 init, newargv = 0x556504b3c240 newargc=9

Fuse and nesting options are checked for the unprivileged Deb10 container running on pve-manager/6.4-13/9f411e79 (running kernel: 5.4.128-1-pve) .

Any hints?

 

[Kiotsoukis]

Since fuse runs in user space, shouldn't that work in LXCs or is there a general issue with LXCs?

 

[Kiotsoukis]

Since fuse runs in user space, shouldn't that work in LXCs or is there a general issue with LXCs?

Ok, doesn't work in an VM too. Funnily I can mount the ceph fs in the VM without problems, just ceph-fuse doesn't work. Debug only shows

auth_reply(proto 2 -13 (13) Permission denied) v1 ==== 24+0+0

although the keyring contains the correct secret.

 

[werty]

I had the same line in debug. This seems to be caused by old unpatched clients that didn't mitigate this issue, and if the cluster set the auth_allow_insecure_global_id_reclaim variable to false. See here for more details. When accessing the same cluster with a newer client (ceph-common 2:14.2.11-147) I got the error monclient(hunting): handle_auth_bad_method server allowed_nmethods [2] but i only support [2,1]

 

[Kiotsoukis]

Oh, sorry, thought this wasn't interesting for anyone. Getting the current ceph-fuse from the ceph repositories solves the problem. The one in the debian repositories is just too old.