Open gillespi314 opened 2 weeks ago
I have been thinking about this. I think a compromise between https://github.com/fleetdm/fleet/pull/19906 and the current state would be to JOIN
with host_mdm
in the queries that determine connected_to_fleet
and assume it's not connected if mdm is off
QA Notes:
Ran thru several different scenarios (similar to QA steps for #19951) including removing the mdm profile while fleetd was still installed. Looking at various tables in the DB (nano_enrollments
, host_mdm
) things are looking good as well as in the UI. I noticed another bug that showed the host was never
added to fleet after manually removing the profiles but that should be fixed with #20059
Bringing over some additional info from Slack and examining the API responses from Noah's "My device" page:
Sarah Gillespie: Have you done anything unusual with this device (maybe erased all content and settings) that could explain why MDM is turned off? Fleet still thinks that you’re
connected_to_fleet
, which means that there is still an active nanoMDM enrollment record for this device in the Dogfood DB. This can happen when Fleet/nano never receives the “Checkout” message from the device that the MDM protocol specifies as the MDM-unenrollment signal.Noah Talerman: I turned MDM off manually (removed the enrollment profile)
Oh interesting. Sounds like a bug?
Sarah Gillespie: We saw something related to this when QA’ing recent changes. My understanding from Roberto is currently an expected side-effect of how the
connected_to_fleet
was defined (it is connected if there is an entry for the device in nano) and the limitations of the MDM protocol checkout flow (we only know to delete nano entries when we receive a checkout message). To handle this case, we’ll need some additional product specification on what to do about stale/orphaned entries. But it is tricky to distinguish between different scenarios and to define the edges. I’ll move this convo to the GitHub issue.Originally posted by @gillespi314 in https://github.com/fleetdm/fleet/issues/19951#issuecomment-2195183767