VSX switch reboot
After a VSX switch reboots, it has no entries for ARP, MAC, and routes. If downstream VSX LAG ports are activated before all this information is relearned, traffic is dropped. To avoid a traffic drop, VSX LAGs on the rebooted switch stay down until the restoration of LACP, MAC, and ARP databases.
The learning process for the VSX LAGs has two phases:
Initial sync phase: The LACP states, MAC address table, and ARP table are downloaded from the forwarding switch to the freshly rebooted switch.
Link-up delay phase: The downloaded entries are installed into the ASIC. Router adjacencies with core nodes and learned upstream routes are also established.
The link-up delay phase is configurable with the
linkup-delay-timer <DELAY-TIMER>
command. The default value is 180 seconds. Set the link-up delay timer to the maximum value of 600 seconds for a network with many MAC addresses, a large ARP table, or a large routing table.
When both VSX switches reboot, the link-up delay timer is not used because both switches are trying to relearn the LACP states, MAC address table, and ARP table.
To get upstream router adjacencies established during the link-up delay, the upstream LAGs have to be excluded from the scope of the link-up delay. Run the
linkup-delay-timer exclude lag-list
<LAG-LIST>
for identifying the LAGs for exclusion.
For example, assume that you have a topology similar to the one in
Sample VSX solution topology, the upstream LAGs (LAG 101 and LAG 102), would need to be identified by the
linkup-delay-timer exclude lag-list
<LAG-LIST>
for exclusion before a VSX switch reboot.