Set blocked status for deferred restarts to indicate operator intervention required

[Garrett Neugent]

Currently, when a restart is deferred, it shows up as active and idle in juju status. However, restarts require operator intervention to resolve, and so the charm hook should set this to a blocked status as well. This applies to all charms with enable-auto-restarts, not just ovn-chassis.

Thanks!

[Andrea Ieri (aieri)]

On top of setting status to blocked or notifying the operator via juju status in some other way, it would probably be useful to also set up nrpe alerting. After all, you eventually do want to apply any pending restart or hook so it would seem appropriate to treat a deferred restart as a type of "error" condition.

[Felipe Reyes (freyes)]

On Mon, 2022-05-30 at 23:14 +0000, Andrea Ieri wrote:
> On top of setting status to blocked or notifying the operator via juju
> status in some other way, it would probably be useful to also set up
> nrpe alerting. After all, you eventually do want to apply any pending
> restart or hook so it would seem appropriate to treat a deferred restart
> as a type of "error" condition.
>
this is one of those scenarios where a nagios check is too strict (OK or not),
so proably the check should encapsulate some logic around the number of deferred
restarts and the age of those events to transition from OK -> Warning ->
Critical.