Wording around osd-devices option can be improved
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
OpenStack Charms Deployment Guide |
Fix Released
|
Low
|
Peter Matulis |
Bug Description
[x] This doc is inaccurate in this way:
the first juju deploy command for the ceph-osd nodes refers to the ceph-osd.yaml file with the osd-devices /dev/sda /dev/sdb /dev/sdc /dev/sdd. As I tried this with /dev/sda, my deployment crashed. I deleted /dev/sda in the yaml file (because of this is the os-device in my case), and everything went well. As I saw, in the documentation for the YOGA-Release, the /dev/sda hadn't been named.
My four machines for ceph-osd have five devices: sda for os, sdb, sdc, sdd and sde for ceph-osd. Should there be a hint, to take care and have a look into MAAS onto the storage of the nodes?
-------
Release: on 2023-02-17 19:50:01
SHA: 1e3c10016f60142
Source: https:/
URL: https:/
Changed in charm-deployment-guide: | |
status: | New → Triaged |
summary: |
- Install OpenStack in charm-deployment-guide + Wording around osd-devices option can be improved |
Hi Robert,
The Ceph charms are designed to not clobber a device that is currently in use so it sounds like you may have hit an important software bug. If you could dig into that it would be appreciated. You would probably file that bug against the ceph-osd charm. Its bug tracker is here:
https:/ /bugs.launchpad .net/charm- ceph-osd/ +filebug
I will still include a note to have the user check the MAAS nodes as you suggested.