Closed loganr-w closed 4 years ago
Setting SCHEDULER_MAX_RECONNECT
to a higher number (say 30) should fix the issue where the framework is going inactive. We set it to 30 and trigged a leader election several times and the framework did not going inactive. We set it to 1, triggered a leader election and the framework went inactive. Hopefully it works.
Checklist
manage.py test
passesAffected app(s)
Scale Scheduler
Description of change
Moved the hard-coded value for max connections in the scheduler to an environment variable.
The hope is that by increasing this number that when a leader election happens that takes an extended amount of time, the framework will not be marked inactive.