Description

This article will provide some useful and general troubleshooting for MC-LAG.

Symptoms

While you are configuring MC-LAG it may be a challenge to troubleshoot and find possible configuration mismatches.

Solution

Here are some troubleshooting steps learned during the last few years that may help you with MC-LAG.

1. Check to make sure the 'set switch-options service-id <service-id value>' command is set on both nodes to the same value.
2. Make sure the MC-LAG nodes are not part of a virtual chassis (VC). MC-LAG is only supported between standalone devices as of 6/22/2015.
3. If an IP address on an mc-ae IRB interface on the peer device needs to be pinged, then set the following command: 

  set interfaces irb unit <unit number> family inet address <IP and mask> arp <IP address of peer IRB> l2-interface <ICL LAG interface> mac <mac address of peer IRB interface> 

4. If VRRP and mcae-mac-synchronize are configured at the same time, you may need to reboot the devices to recover. See PR 1161830.

From the note here, it seems we can't have both VRRP and mcae-mac-synchronize configured: 

"Note: On QFX Series switches, you cannot configure both VRRP over IRB and MAC synchronization, because processing MAC addresses might not work." 

http://www.juniper.net/documentation/en_US/junos13.3/topics/concept/lag-multichassis--guidelines.html 

5. Restart ICCP? 

6. Look for any other VC-related config because MC-LAG is not supported on VC. 

7. If there is an issue with mac learning on non mc-lag vlans between the mc-lag peers, make sure 

the vlans are not trunked on the ICL. They should be trunked on their own separate LAG. 

8. Make sure the IP address of the MC-LAG peer in the statement 'set protocols iccp peer <IP address of peer> is reachable. 

9. If ICCP is not up, make sure that the ICCP vlan is not configured on the mc-ae interfaces. 

10. If STP is configured along with MC-LAG, then the Juniper device must be the root bridge: 

https://www.juniper.net/documentation/en_US/junos/topics/concept/mc-lag-feature-summary-best-practices.html 

"With RL2GP, you can configure two edge MC-LAG nodes with the same STP virtual root ID. The virtual root ID must be superior to all bridges in the downstream network". 

11. Run the following commands: 

# run show multi-chassis mc-lag configuration-consistency | match FAIL 
# run show multi-chassis mc-lag configuration-consistency 

Considerations:
  • "set interface <interface-name> vlan-tagging/flexible-vlan-tagging" on trunk ports to increase the MTU to 1518/1522. 
  • If this knob is not there then packet size of larger than 1514 will get dropped. 
  • "set interface irb arp-l2-validate" to ensure there is no mistmach in arp and mac. Only valid on EX9200. Not valid on Broadcom chipsets. 
  • "set forwarding-options dhcp-relay forward-snooped-clients <>" when dhcp-relay is configured. By default, dhcp packets are snooped 
  • if dhcp-relay is configured and default action is dropped if it not part of the relay. 
  • "set forwarding-options dhcp-relay route-suppression destination". This is to ensure that arp entry is not learnt via jdhcpd on ICL link. 
  • "set protocols igmp-snooping vlan all interface <> multicast-router-interface". Always configure ICL as mrouter-interface for multicast; otherwise ICL will not forward multicast traffic. 
  • "set multicast-snooping-options multichassis-lag-replicate-state" Multicast-replicate-state is required for IGMP snooping to work. 

 

  • Check chassis configuration to make sure the device supports aggregate ethernet interfaces:

user@router>show configuration chassis                      
aggregated-devices {
    ethernet {
        device-count 200;
  • Make sure the physical interface is correctly assigned to LAG:

user@router> show configuration interfaces xe-0/0/0
description test;
gigether-options {
    802.3ad ae50;
}
  • Both LAG peers should have unique chassis ID.
  • Both MC-LAG peers should have the same mode. The two available modes are active-active or active-standby
  • Since these two MC-LAG peers looks the same device from a remote end perspective, then both MC-LAG peers should have the same LACP ID.
  • Both MC-LAG peers should have the same LACP admin key.
  • The switch service ID is used to synchronize applications, IGMP, ARP, and MAC learning across MC-LAG members. Therefore, we define the same service-id on both peers.

Tips:

  • Do not forget to specify the child interfaces that belong to the MC-AE interfaces.
  • The MC-LAG will not become active if there is no unit configuration for the AE.
  • When the interface of an active -active MAC-LAG is enabled for ethernet switching, the VLAN needs to be added to the ICL as well.
  • You can reuse most settings for a second MC_AE between MC-LAG peers. Only the system ID, admin key and mc-as-id need to differ on a per MC-AE basis.
  • Certain MC-LAG deployments involving active routeing-protocols between the MC-LAG peers may require static ARP entries.
  • If the remote end does not support LACP, use the "force-up" option to make sure that MC-AE interfaces become operational

Modification History

06-16-2023:KB Created

Related Information

Best Practices and Usage Notes

Advanced MC-LAG Concepts

MC-LAG Concepts

MC-LAG Examples