Trove services (api, conductor, taskmanager) do not start if the sql server is not enabled
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
OpenStack DBaaS (Trove) |
Confirmed
|
Medium
|
Unassigned |
Bug Description
When the sql server used in the sql_connect parameter is not active, the services do not start (tested with mariadb, but I think it does not matter which one is used).
I'm pointing this out because other services from other OpenStack project are able to start when the db is not available (checked nova-*, glance-*, cinder-* services; notable exception: neutron-server does not start). They return an error when the db is not available, but they recover and start working again when the db server is restarted.
If the current one (hard dependency on the db also for _starting_ the server) is the expected behavior of trove services, at least it should be documented to provide guidance to packagers.
Changed in trove: | |
status: | New → Confirmed |
importance: | Undecided → Medium |
Moreover, it is not always possible to solve this with dependencies, as the database server could live on a different host. The services should really be able to start even when the backend db is not reachable.