On Sat, 2019-10-19 at 10:36 +0530, Rishabh Dave wrote:
Hi,
I tried mounting CephFS with kernel driver on a vstart cluster on
master branch (latest commit SHA:
d33c281b6437523a66d7802a39514f1ae74ec8e7) without secret key, but I
was unsuccessful.Following is a copy of stdout while I was trying to
mount. The first mount failed because I picked the wrong port.
However, the second attempt (with key) was successful and the third
(without key) wasn't.
build$ sudo mount -t ceph 192.168.0.218:40112:/ /mnt/kcephfs -o
name=admin,secret=AQDrjqpdy0fGKhAATIRQrdPhXB/uIi+86xuijQ==
^C
build$ sudo mount -t ceph 192.168.0.218:40113:/ /mnt/kcephfs -o
name=admin,secret=AQDrjqpdy0fGKhAATIRQrdPhXB/uIi+86xuijQ==
build$ sudo umount /mnt/kcephfs/
build$ sudo mount -t ceph 192.168.0.218:40113:/ /mnt/kcephfs -o name=admin
mount: /mnt/kcephfs: wrong fs type, bad option, bad superblock on
192.168.0.218:40113:/, missing codepage or helper program, or other
error.
build$ dmesg | tail
[ 806.561086] libceph: mon0 192.168.0.218:40112 socket closed (con
state CONNECTING)
[ 810.770148] libceph: mon0 192.168.0.218:40113 session established
[ 810.772603] libceph: client4275 fsid 54b1853a-1a08-482d-baf4-644eec15e830
[ 822.452439] libceph: no secret set (for auth_x protocol)
[ 822.452443] libceph: error -22 on auth protocol 2 init
build$
Just to make sure, I tried a fourth time with key to -
build$ mount -t ceph 192.168.0.218:40113:/ /mnt/kcephfs -o
name=admin,secret=AQDrjqpdy0fGKhAATIRQrdPhXB/uIi+86xuijQ==
build$ $ mount | grep kcephfs
192.168.0.218:40113:/ on /mnt/kcephfs type ceph
(rw,relatime,name=admin,secret=<hidden>,acl)
Thinking that mount.ceph helper might be looking for file
`ceph.client.admin.keyring`, I copied the admin keyring in a file, and
placed it build/ as well as in /etc/ceph. However, that too didn't
help. I've copied shell output for mount commands and contents keyring
files here, in case that helps -
https://paste.fedoraproject.org/paste/YbFY235S3DaEryje9HDPAw.
Now that I look again, I think you're probably just getting the wrong
mount helper here.
If it exists, /bin/mount will exec the /sbin/mount.ceph binary to do a
'-t ceph' mount. If you're running this on the box where you built ceph,
then it's probably using whatever mount.ceph binary happens to be
installed in /sbin rather than the one you built.
You might want to try turning /sbin/mount.ceph into a symlink to the
mount.ceph binary under your build dir, or just copy it into place.
--
Jeff Layton <jlayton(a)redhat.com>