Expected High Availability failover time?

What’s the expected time for a HA master to slave failover to occur?

The manual says the failover should happen after 3 seconds but in my testing the failover is taking ~10 seconds.
Is this expected? If so, can it be improved?

@JasonHilton, may I know how you conducted the failover testing?

As I have tested with our devices (Balance 305, FW8.2.1), when the Master unit LAN interface is disconnected, it will switch to Slave mode in 3 seconds, as per the user manual stated.

Do take note that when the Slave unit take over as Master, the WAN interface will initialize/negotiate the connection, before client traffic can be forwarded. Maybe you can consider this option (highlighted) to help shorten the downtime.

@WeiMing Thanks for the quick reply!
I have that enabled as well.
For testing, I had a ping to google.com running. When disconnecting the master, there were 9 failed pings before the slave took over.

I am following the similar setup in my lab test:
image.png

  1. When I have Config Sync. with the Master (w/o the “Establish Connections in Slave Role”), I have about 5 PING timeout to 8.8.8.8 during the HA failover (disconnect LAN).

  2. If I enabled the “Establish Connections in Slave Role”, I get average 2-3 PING timeout to 8.8.8.8 during the HA failover.

[NOTE]
I am using 2x B305 (HW2) running firmware 8.2.1GA, with DHCP/Static IP WAN.

Hope this info helps. :slight_smile:

1 Like

@WeiMing Thanks for checking. I have the same configuration as above. Perhaps this is a SDXP issue? Screen recording of issue


May I know the brand of the switch that use for the setup ? Sometime it will be influence by the switch port security settings (STP, Fast port) that may delayed the fail-over.

Hi @sitloongs

It’s a Peplink 48 port, with the SDXPs connected to SFP 1&2.