How Failover & Load Balancing Work

Topics:
•
WAN Interface Failure
•
WAN Interface Recovery

WAN Interface Failure

This is what FLB does when a WAN interface failure had been detected (linkDown or probing-failure or no-IP-settings):

1
Graceful shutdown of the interface (call the stop API, if one is provided; for example, pppoe-stop, dialup-stop).
2
Trigger the disabling of routes associated with the failed interface (except for the ones marked do not disable on link down).
3
Flush dynamic ARP entries using the failed interface.
4
Flush the cache entries using the failed interface as the outbound interface.
5
Update the WAN default route to point to an alternate WAN, if available. Update status data (this is part of recovery procedure).
a
Address Objects used by other apps such as CASS gets updated as well.
b
Security Services depend on this for failover capability.
6
Notify interested parties (VPN, BWM, CASS, DDNS, DNS).
7
Actively monitor status of failed interface, attempt recovery such as restarting WAN connection (call the start API, if provided; for example, pppoe-start, dial-start).

WAN Interface Recovery

This is what FLB does when a WAN interface recovery had been detected (linkUp or probing-success or IP-change):

1
On linkUp, jump-start the interface connection (call the start API, if provided; for example, pppoe-start, dial-start). In most cases, this would be in a connected state already, but if it is not, FLB attempts to push it to start. It may do a graceful shutdown and restart if a hung condition is detected (timer based).
2
When connectivity is confirmed (simple linkUp or probing), trigger enabling of routes associated with the interface.
3
Add ARP entries (if any are needed).
a
Send out unsolicited ARP response (for interface) to update neighboring devices.
4
If needed, update the WAN default route (for example, preempt) to use the best available WAN. Update status data.
a
Address Objects used by other apps such as CASS gets updated as well.
b
Security Services depend on this for failover capability.
5
Notify interested parties (VPN, BWM, CASS, DDNS, DNS).
6
Continue monitoring status of interface.