Slow VPLS convergence after MC-LAG failover
A service provider aggregates 300 business customers through two Nokia 7750 SR routers, AGG-1 and AGG-2, configured as an MC-LAG pair. Access switches are dual-homed with LACP. AGG-1 is active and AGG-2 is standby. Layer 3 services use BFD and fail over in under 200 ms. A VPLS VLAN carrying IPTV multicast has a 25-second outage whenever AGG-1's active link fails or AGG-1 reloads. During the outage, multicast receivers freeze and MAC addresses remain pointing to AGG-1. The MC-LAG uses LACP with default timers; STP is disabled on the access switches; IGMP snooping is enabled on the VPLS. The provider wants sub-second L2 convergence and no duplicate multicast.
1. Why does L3 fail over in under 200 ms while the VPLS VLAN takes 25 seconds? Identify the L2 mechanisms involved.
Read the worked answer
L3 uses BFD, which detects the failed link in milliseconds and withdraws the route. L2 depends on LACP and MAC learning. With default slow LACP timers, the access switch may take tens of seconds to declare the link down. During that time, traffic continues to hash to AGG-1. Even after LACP detects failure, the VPLS core still has MAC addresses learned via AGG-1 and IGMP snooping state for the IPTV groups. Because STP is disabled on the access switches, there is no topology change to trigger a flush. The 25-second outage matches slow LACP plus stale MAC and multicast state.
2. Recommend SR OS MC-LAG and VPLS settings to achieve sub-second L2 convergence and avoid duplicate multicast.
Read the worked answer
On the MC-LAG, enable LACP fast timers (1 second) and configure consistent LACP keys. Enable standby signaling so AGG-2 knows it is standby and can take over cleanly. Enable MAC flush on MC-LAG failure and configure the VPLS to send flush messages to remote PEs when the active link fails. For multicast, enable IGMP snooping with a querier and fast leave, and ensure the snooping entries are flushed on failover. If duplicate multicast is a concern, use IGMP snooping proxy on the aggregation pair. Avoid disabling STP globally without an alternative loop-prevention design; with dual-homed access and MC-LAG, the access switch should not create a loop, but verify.
3. How would you verify failover without disrupting customers?
Read the worked answer
Use a maintenance window with a test VLAN and a multicast receiver. Generate traffic, fail AGG-1's active link, and measure LACP timeout, MAC flush counters, IGMP join latency, and packet loss. Check that AGG-2 becomes active and that no duplicate streams arrive. Repeat by reloading AGG-1. Monitor MC-LAG status, LACP state, VPLS MAC counts, and IGMP snooping entries.