Improve Database reconnect logic
Currently, when kea-dhcp4/6 servers lose connectivity to any of their backends (lease, host, or CB), the reconnect logic attempts to reconnect to all of them, whether they were lost or not. This is not the most efficient or flexible thing to do.
The reconnect logic is here:
ControlledDhcpv*Srv::dbReconnect(ReconnectCtlPtr db_reconnect_ctl)
This function is the lost db callback function and is invoked by the DatabaseConnection that suffers the failure. To improve behavior we will likely need more information passed in via db_reconnect_ctl, such that the above function can identify which backend has been lost and reconnect only that one.
We have a support customer that suggests we might want the ability to treat the loss of CB as non-fatal: