As far as I recall, 'Create volume and boot instance from it' pretty frequently failed on instance deletion if the test is run after RabbitMQ failover. I mean, that happened pretty frequently on custom destructive tests for me in previous releases (7.0, 8.0).
The reason is that after RabbitMQ failover OpenStack needs a 'warm-up' load to clean broken connections. Until that load passes, OpenStack lags a little as it clears broken connections and establishes new one on the fly. So either the failed test serves as a 'warm-up' for later tests (the test comes first in the list). Or it just executes almost hitting the limit under normal conditions and lags beyond the limit after failover.
As far as I recall, 'Create volume and boot instance from it' pretty frequently failed on instance deletion if the test is run after RabbitMQ failover. I mean, that happened pretty frequently on custom destructive tests for me in previous releases (7.0, 8.0).
The reason is that after RabbitMQ failover OpenStack needs a 'warm-up' load to clean broken connections. Until that load passes, OpenStack lags a little as it clears broken connections and establishes new one on the fly. So either the failed test serves as a 'warm-up' for later tests (the test comes first in the list). Or it just executes almost hitting the limit under normal conditions and lags beyond the limit after failover.