Comment 9 for bug 1698154

Edward Hope-Morley (hopem) wrote :

I just noticed this bug - https://bugs.launchpad.net/charm-ceph-osd/+bug/1746118 - which is similar but not the same as this one but I believe they can/should both leverage the same fix but with a slight adjustment. 1746118 does a udevadm settle after having formatted and started osds to ensure they don't get reformatted on a subsequent hook fire but before all the new osds have been mounted. The trouble is that the settle is done after the formatting where we should really be doing it before as well (maybe just if osd-reformat=True) so as to ensure that there are not devices still waiting to be mounted before the code checks for unmounted devices.