On 08/02/2021 16:52, Kenneth Waegeman wrote:
Hi Eugen, all,
Thanks for sharing your results! Since we have multiple clusters and
clusters with +500 OSDs, this solution is not feasible for us.
In the meantime I created an issue for this :
https://tracker.ceph.com/issues/49159
Hi all,
For those who would have same/similar issues/questions, the ticket has
been updated.
it actually breaks down in two parts:
- ceph-volume documentation
(https://docs.ceph.com/en/latest/ceph-volume/lvm/activate/#activate
<https://docs.ceph.com/en/latest/ceph-volume/lvm/activate/#activate>)
notes that activate means:
'This activation process enables a systemd unit that persists the OSD ID
and its UUID (also called fsid in Ceph CLI tools), so that at boot time
it can understand what OSD is enabled and needs to be mounted.'
-> This is not true/does not work for use with cephadm, ceph-volume
can't make the osd directories/files like unit.run (yet) for osds that
should run with cephadm
- there is yet no way (documented) that existing OSD disks could be
discovered by cephadm/ceph orch on reinstalling a node like it used to
be with running ceph-volume activate --all. The workaround I see for now
is running
ceph-volume activate --all
for id in `ls -1 /var/lib/ceph/osd`; do echo cephadm adopt --style legacy --name ${id/ceph-/osd.}; done
This removes the ceph-volume units again and creates the cephadm ones :)
As pointed out by Sebastian Wagner: 'Please verify that the container
image used is consistent across the cluster after running the adoption
process.'
And thanks @Sebastian for making 'cephadm ceph-volume activate' a
feature request!
Kenneth
We would need this especially to migrate/reinstall all our clusters to
Rhel8 (without destroying/recreating all osd disks), so I really hope
there is another solution :)
Thanks again!
Kenneth
On 05/02/2021 16:11, Eugen Block wrote:
Hi Kenneth,
I managed to succeed with this just now. It's a lab environment and
the OSDs are not encrypted but I was able to get the OSDs up again.
The ceph-volume commands also worked (just activation didn't) so I
had the required information about those OSDs.
What I did was
- collect the OSD data (fsid, keyring)
- create directories for osd daemons under
/var/lib/ceph/<CEPH_UUID>/osd.<ID>
- note that the directory with the ceph uuid already existed since
the crash container had been created after bringing the node back
into the cluster
- creating the content for that OSD by copying the required files
from a different host and changed the contents of
- fsid
- keyring
- whoami
- unit.run
- unit.poststop
- created the symlinks to the OSD devices:
- ln -s /dev/ceph-<VG>/osd-block-<LV> block
- ln -s /dev/ceph-<VG>/osd-block-<LV> block.db
- changed ownership to ceph
- chown -R ceph.ceph /var/lib/ceph/<UUID>/osd.<ID>/
- started the systemd unit
- systemctl start ceph-<CEPH_UUID>@osd.<ID>.service
I repeated this for all OSDs on that host, now all OSDs are online
and the cluster is happy. I'm not sure what else is necessary in case
of encrypted OSDs, but maybe this procedure helps you.
I don't know if there's a smoother or even automated way, I don't
think there currently is. Maybe someone is working on it though.
Regards,
Eugen
Zitat von Kenneth Waegeman <kenneth.waegeman@xxxxxxxx>:
Hi all,
I'm running a 15.2.8 cluster using ceph orch with all daemons
adopted to cephadm.
I tried reinstall an OSD node. Is there a way to make ceph
orch/cephadm activate the devices on this node again, ideally
automatically?
I tried running `cephadm ceph-volume -- lvm activate --all` but this
has an error related to dmcrypt:
[root@osd2803 ~]# cephadm ceph-volume -- lvm activate --all
Using recent ceph image docker.io/ceph/ceph:v15
/usr/bin/podman:stderr --> Activating OSD ID 0 FSID
697698fd-3fa0-480f-807b-68492bd292bf
/usr/bin/podman:stderr Running command: /usr/bin/mount -t tmpfs
tmpfs /var/lib/ceph/osd/ceph-0
/usr/bin/podman:stderr Running command: /usr/bin/ceph-authtool
/var/lib/ceph/osd/ceph-0/lockbox.keyring --create-keyring --name
client.osd-lockbox.697698fd-3fa0-480f-807b-68492bd292bf --add-key
AQAy7Bdg0jQsBhAAj0gcteTEbcpwNNvMGZqTTg==
/usr/bin/podman:stderr stdout: creating
/var/lib/ceph/osd/ceph-0/lockbox.keyring
/usr/bin/podman:stderr added entity
client.osd-lockbox.697698fd-3fa0-480f-807b-68492bd292bf
auth(key=AQAy7Bdg0jQsBhAAj0gcteTEbcpwNNvMGZqTTg==)
/usr/bin/podman:stderr Running command: /usr/bin/chown -R ceph:ceph
/var/lib/ceph/osd/ceph-0/lockbox.keyring
/usr/bin/podman:stderr Running command: /usr/bin/ceph --cluster
ceph --name client.osd-lockbox.697698fd-3fa0-480f-807b-68492bd292bf
--keyring /var/lib/ceph/osd/ceph-0/lockbox.keyring config-key get
dm-crypt/osd/697698fd-3fa0-480f-807b-68492bd292bf/luks
/usr/bin/podman:stderr stderr: Error initializing cluster client:
ObjectNotFound('RADOS object not found (error calling
conf_read_file)',)
/usr/bin/podman:stderr --> RuntimeError: Unable to retrieve
dmcrypt secret
Traceback (most recent call last):
File "/usr/sbin/cephadm", line 6111, in <module>
r = args.func()
File "/usr/sbin/cephadm", line 1322, in _infer_fsid
return func()
File "/usr/sbin/cephadm", line 1381, in _infer_image
return func()
File "/usr/sbin/cephadm", line 3611, in command_ceph_volume
out, err, code = call_throws(c.run_cmd(), verbose=True)
File "/usr/sbin/cephadm", line 1060, in call_throws
raise RuntimeError('Failed command: %s' % ' '.join(command))
RuntimeError: Failed command: /usr/bin/podman run --rm --ipc=host
--net=host --entrypoint /usr/sbin/ceph-volume --privileged
--group-add=disk -e CONTAINER_IMAGE=docker.io/ceph/ceph:v15 -e
NODE_NAME=osd2803.banette.os -v /dev:/dev -v /run/udev:/run/udev -v
/sys:/sys -v /run/lvm:/run/lvm -v /run/lock/lvm:/run/lock/lvm
docker.io/ceph/ceph:v15 lvm activate --all
The OSDs are encrypted indeed. `cephadm ceph-volume lvm list` and
`cephadm shell ceph -s` run just fine, and if I run ceph-volume
directly, the same command works, but then of course the daemons are
started in the legacy way again, not in containers.
Is there another way trough the 'ceph orch' to achieve this? Or if
`cephadm ceph-volume -- lvm activate --all` would be the way to go
here, I'm probably seeing a bug here ?
Thanks!!
Kenneth
_______________________________________________
ceph-users mailing list -- ceph-users@xxxxxxx
To unsubscribe send an email to ceph-users-leave@xxxxxxx
_______________________________________________
ceph-users mailing list -- ceph-users@xxxxxxx
To unsubscribe send an email to ceph-users-leave@xxxxxxx
_______________________________________________
ceph-users mailing list -- ceph-users@xxxxxxx
To unsubscribe send an email to ceph-users-leave@xxxxxxx
