Link-up delay

When a VSX device is rebooted, it has no entries for MAC, ARP, routes. If downstream VSX LAG ports are activated before the information is relearned, traffic is dropped. To avoid a traffic drop, VSX LAGs on the rebooted device stay down until the restore of LACP, MAC, ARP, and MSTP databases.

The learning process has two phases:

  • Initial synchronization phase:
    • This phase is the download phase where the rebooted node learns all the LACP+MAC+ARP+STP database entries from its VSX peer through ISLP.

    • The initial synchronization timer, which is not configurable, is the required time to download the database information from the peer.

  • Link-up delay phase:

    • This phase is the duration for:

      • Installing the downloaded entries to the ASIC.

      • Establishing router adjacencies with core nodes and learning upstream routes.

    • The link-up delay timer default value is 180 seconds.

    • Depending on the network size, ARP/routing tables size, you might be required to set the timer to a higher value (maximum 600 seconds).

When both VSX devices reboot, the link-up delay timer is not used.

To get upstream router adjacencies established during the link-up delay, the upstream LAG (for example LAG 101) has to be excluded from the scope of the link-up delay. Even if the upstream VSX node is not excluded from the link-up delay timer, OSPFv2/OSPFv3 neighborship forms, when active-forwarding is enabled on a VLAN. While the link-up delay timer is running, all SVIs that contain VSX LAG members are kept in a pseudo-shutdown state.