When charm checks for cluster_initialized, the property checks for cluster
metadata on any of the units. Case the first unit is unreachable (e.g. freeze test)
the call hangs, impairing charm operation for all units (not only the unreachable one).
And that's when the freeze test hangs, specially on k8s, i.e. when the first unit myqsld
is the one being stopped.
Issue
When charm checks for
cluster_initialized
, the property checks for cluster metadata on any of the units. Case the first unit is unreachable (e.g. freeze test) the call hangs, impairing charm operation for all units (not only the unreachable one).And that's when the freeze test hangs, specially on k8s, i.e. when the first unit myqsld is the one being stopped.
Solution
Adds a 10s (scientific) timeout to the call.