Kolla ceph_osd bootstrap doesn't work with NVMe drives

[Marius L]

In the linux systems, the NVMe drives are named like /dev/nvme0n1 and partitions are appended with "p", like /dev/nvme0n1p1.

After tagging the nvme devices with "KOLLA_CEPH_OSD_BOOTSTRAP_BS", the bootstrap_ceph_osd container create wrong partition names: "/dev/nvme0n11" instead of "/dev/nvme0n1p1", resulting the following error:

"Problem opening /dev/nvme0n11 for reading! Error is 2.
The specified file does not exist!
Problem opening '' for writing! Program will now terminate.
Warning! MBR not overwritten! Error is 2!"

[Eric Miller (erickmiller)]

In case this helps someone, I modified the current master files to clean up a few things and allow for nvme drives to be used. Not much has changed over the past year or so, so the master version was plenty good. We use Rocky currently and these changes work fine. I have only tested a bluestore deployment, where we have both SATA SSDs and HDDs, and now NVMe SSDs (all assigned to their respective Ceph device classes).

Note that we tried to get LVM working, but too many changes were required, so these changes only allow a single OSD to use an NVMe device.

Maybe someone can review and merge the changes into master?

The two files that need adjustment are:
/usr/share/kolla/docker/ceph/ceph-osd/extend_start.sh
/usr/share/kolla/docker/kolla-toolbox/find_disks.py

See the attached files.

Eric

[Michal Nasiadka (mnasiadka)]

Addressed in this PR: https://review.opendev.org/#/c/688328