WAN2 latency spiking when WAN1 is down.
-
I have configured WAN1 with a custom monitoring IP (8.8.8.8) since the gateway doesn't respond to ICMP Echo requests. Whenever WAN1 PPPoE goes down or is manually disconnected, WAN2 repeatedly spikes to latencies of 1000ms at regular intervals. It is not packet loss. The packets are delayed by 1000ms. This is evident in games when it responds uniformly with a constant delay of 1000ms. In the Interfaces status screen, it shows the WAN2 gateway with a latency of 1000ms. WAN1 is the default gateway of the pfSense box.
-
That sounds like one of the weird edge case bugs in apinger that was fixed in 2.2. In some circumstances with a down gateway, it'd report inaccurate latency numbers. If it's replicable, you might want to try a 2.2 snapshot to see if it's an issue there as well.
-
It's not just "reporting" a higher latency. The actual latency is higher. My pings are consistently 500ms or 1000ms higher when I ping any address or play games. It literally takes 1 second more to respond. If I disconnect and reconnect WAN2 it's normal for a few seconds than the ping increases again until I restore WAN1.
-
Oh, that's definitely not that. Have traffic shaping or limiters configured?
-
@cmb:
Oh, that's definitely not that. Have traffic shaping or limiters configured?
No. It's a fresh installation with basic firewall rules and the WAN2 gateway specified for some source ports as seen here: https://forum.pfsense.org/index.php?topic=83526
The latency to the WAN2 default gateway IP address from PPPoE itself is 1000ms.
-
Sounds like what would happen if you're maxing out your upload bandwidth on that connection, or if that connection has issues in general. Is it that high with the connection idle?
-
@cmb:
Sounds like what would happen if you're maxing out your upload bandwidth on that connection, or if that connection has issues in general. Is it that high with the connection idle?
It's idle. No further traffic goes over WAN2 when WAN1 is down. It's not a failover. Only UDP source port 27005 and some game client UDP destination ports are permanently sent through WAN2. The increased latency is always 500 or 999 ms + the actual route latency to be exact.
Even when WAN1 is up and I'm playing a game, the game is going through WAN2. During this time the WAN2 latency is normal. If WAN1 fails, the latency goes up shortly.
A related problem as mentioned in the linked thread is when WAN2 is down, the above rules don't fall back to WAN1.