You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
When a user creates a database based on a template a FATAL error will
be generated, which causes the scheduler to restart. The same happens
if a FATAL error is generate in any other way, e.g., if the database
does not exist.
If you use an actual template database (like template0) this can
cause endless errors and retries wasting bgw slots. Hence, we avoid
starting the bgw scheduler for databases that does not exist and
databases we cannot connect to.
In addition to when creating a new database, a re-starting scheduler
might generate errors if the database has gone away or if permissions
for the database changed. This will trigger an endlessly restarting
scheduler.
We fix this issue by overriding connection block when connecting to the
database and perform the check ourselves to avoid a fatal error, and
capturing any non-fatal error that can be generated in the scheduler
startup phase and instead report the error ourselves and exiting with
exit code 0 (which means the scheduler will not restart).
If an error occurs inside the main scheduler process, this will still
trigger a restart.
Co-authored-by: Fabrízio de Royes Mello <[email protected]>
Co-authored-by: Mats Kindahl <[email protected]>
0 commit comments