[21.05] fc-ceph: improve osd safety check UX

[osnyx]

In accordance to everyday operations usage, reduce the strictness of the
default safety check when destroying OSDs. By default, we use
ok-to-stop which allows degraded data redundancy as long as the
cluster stays IO-operational.
The old safe-to-destroy check behaviour is available under the new
flag --strict-safety-check.

This overall increases the UX safety of common operations tasks, as we accept temporarily degrading the redundancy for rebuilds as long as data stays available. Previously, we had to disable safety checks altogether, now the default safety check allows these operations but still prevents us from breaking availability.

The check mechanism has also been added to fc-ceph osd deactivate. The reactivate operation does its own checking anyways and is not affected.

@flyingcircusio/release-managers

Release process

Impact: internal only

Changelog:

• fc-ceph:
  • reduce strictness of default osd destruction safety checks
  • rename flags --unsafe-destroy -> --no-safety-check
  • add safety checks for deactivate

Security implications

• Security requirements defined? (WHERE)
  • The UX of our tooling should be designed along our common operation actions. Safety checks should support the operator from accidentally entering dangerous states, while not influenceing common operations too much.
  • no new (known) regressions must be introduced
• Security requirements tested? (EVIDENCE)
  • extended test cases to cover new functionality
  • automated tests still pass
  • manually tested in dev cluster

[ctheune]

Generally looks good with some suggestions.

[ctheune]

Happy with this. However, hydra is red?