2 May
2024
2 May
'24
10 p.m.
1) Is it normal that ceph-volume lvm batch doesn't support partitions for
block.db ?
For some time now, yeah, ceph-volume has been blocking use of partitions
for all device types. It came from
https://github.com/ceph/ceph/commit/9742efa907aa54b3135f5daf73080b7be12534eb
<https://github.com/ceph/ceph/commit/9742efa907aa54b3135f5daf73080b7be12534eb>
a few years ago. I've had discussions with the current ceph-volume
maintainer about whether this actually should be the case or not that have
implied we may want to change this, but that is the current behavior.
2) Is there any way to configure the ceph deployment
to use ceph-volume
lvm create instead of the batch command? Will ceph-volume lvm create
support partitions for block.db ?
Not in a cephadm deployment. We route all the OSD creation through the
batch command. The best I can think of to do right now would be to create
an LV on top of the partition. I think this works because batch will take
LVs. LVs can only be passed using paths though and I think I remember the
format for the paths with LVs not being intuitive. Something like <vg>/<lv>
(e.g. vg_nvme/lv_4)
On Thu, May 2, 2024 at 10:19 AM Yonatan Zaken <yzaken(a)redhat.com> wrote:
> Hi All,
>
> I am using openstack's tripleo wallaby version for deployment of my
> overcloud and using the openstack overcloud ceph deploy command for
> deploying the ceph cluster.
> The Ceph version is Reef release v8.2.2.
>
> My templates/ceph_spec.yaml file contains the following:
>
> data_devices:
> paths:
> - /dev/vdb
> - /dev/vdc
> db_devices:
> paths:
> - /dev/vdd1
> placement:
> hosts:
> - host1
> - host2
> service_id: default_drive_group
> service_name: osd.default_drive_group
> service_type: osd
>
>
> *During deployment of the ceph cluster I noticed it fails when trying to
> execute the following command which is executed in a container (as seen in
> /var/log/ceph/cephadm.log on the storage node):*
> ceph-volume lvm batch --no-auto /dev/vdb /dev/vdc --db-devices /dev/vdd1
> --yes --no-systemd
>
> *with the following error:*
>
> /bin/podman: stderr ceph-volume lvm batch: error: /dev/vdd1 is a
> partition, please pass LVs or raw block devices
>
>
> *My questions are:*
> 1) Is it normal that ceph-volume lvm batch doesn't support partitions for
> block.db ?
2) Is there any way to configure the ceph deployment
to use ceph-volume
lvm create instead of the batch command? Will ceph-volume lvm create
support partitions for block.db ?
> Best Regards,
> Yonatan
> _______________________________________________
> Dev mailing list -- dev(a)ceph.io
> To unsubscribe send an email to dev-leave(a)ceph.io
>