Misconfiguration of bluestore wal/db block size options leads to non-related error message

[Vladimir Grevtsev]

Problem statement:

Charm expects the "bluestore-block-db-size" and "bluestore-block-wal-size" be configured per OSD device, not per unit (it's not clear from config options, here is a review [1] that fixes that).

If this config parameters will be oversized (in my case I specified 480GB for DB and 20GB for WAL, assuming 500GB partition will be used for WAL/DB data per all OSDs; but this parameters was considered to be used per OSD) units will switch to "blocked" status, but with non-related "Non-pristine devices detected" error message - however, charm has failed to execute "lvcreate" CLI command when trying to create WAL/DB partitions.

In this particular case /dev/disk/by-dname/nvme0n1-part2 have 500G overall, but by mistake charm config contained sizing per unit, not per OSD - so this error triggered, and after clearing up and dividing wal/db sizing by OSD count, all is working fine.

Expected output:

Error message like "Unsufficient space on device" or something else; "non-pristine device" appeared here because of first OSD device has been initialized successfully, while others don't and, as I think, it's not related to that case.

Additional information:

juju status: https://pastebin.canonical.com/p/M8YcGgcpcZ/
bundle: https://pastebin.canonical.com/p/4TRwF8hcQJ/
unit logs: https://pastebin.canonical.com/p/rhtK5SMXkC/

[1] https://review.openstack.org/#/c/618527/

[Ryan Beisner (1chb1n)]

The added clarity in the charm config documentation is quite helpful and appreciated (merged 8mo. ago).

I also agree that we can and should make improvements to the UX in the certain scenarios as highlighted above.

This would be a good bug task for engineering onboarding and/or engineering rotations.