I excluded stat replication and replication slots from the tests because
we're probably going to remove those soon anyways.
When we wired this up to haproxy in prod we saw some db conn spikes. Did not impact db load but was making some connection charts look strange. This change will make it so we only perform a real db check once a second. Obviously caching a health check is a strange concept, but our health-checks run every 2 seconds so I think this is a safe change.
I excluded stat replication and replication slots from the tests because we're probably going to remove those soon anyways.
When we wired this up to haproxy in prod we saw some db conn spikes. Did not impact db load but was making some connection charts look strange. This change will make it so we only perform a real db check once a second. Obviously caching a health check is a strange concept, but our health-checks run every 2 seconds so I think this is a safe change.
@benschinn