0

I have a scenario where I have an EX4300 that needs to have redundant links to an MX204 router. My initial thought was to setup LAG with LACP, but I am wondering if this is possible or not. I couldn't find any documentation for setting up LAG between a switch and a router. Then I was thinking maybe VRRP was a better solution. I'm just looking for input from more experienced engineers because I don't have a lot of experience with either of these two applications other than some testing labs a few years ago.

Perhaps there is a better way I'm unaware of, also. What I am trying to accomplish is to have fiber trunks on the EX4300, ports xe-0/2/0 and xe-0/2/1 connected to an MX204 on xe-1/1/0 and xe-1/1/1 respectively, act as one logical link with a gateway IP address configured on the router's two interfaces (or one logical interface). Load balancing would be nice, but it's not necessary.

2
  • 2
    Switch vs. Router doesn't matter as long as both platforms support LACP (which they do). Both devices run Junos, so the same configuration will work for both. Commented May 13, 2021 at 1:36
  • Thanks Jordan, but the thing I'm a bit confused about is how the configuration would look. Right now the interface on the router is advertising a /64 subnet and the switch interface is a trunk. If LAG was implemented, would the switch ae interface still just be a configured as a trunk interface and the ae on the router as the gateway with an IP in the /64? Also, I forgot to ask, are their any limitations with using two routing instances? Could I have say ae.100 in the default routing instance and an ae.200 in another instance?
    – Ben
    Commented May 13, 2021 at 3:17

1 Answer 1

1

LAG aggregates multiple physical links into a single, logical one. It can achieve link-level redundancy and a (possibly limited) increase in link bandwidth.

With very few exceptions, LAG doesn't load balance traffic across the physical links. Instead traffic is distributed using a static scheme. Depending on your hardware and its configuration, that distribution is based on the source and destination MAC addresses (most primitive), the source and destination IP addresses (somewhat usable), or IP addresses and transport-layer port numbers (best).

Statistically, these traffic distribution schemes work - more or less - well with a large number of different flows, but much less so with a small number of flows.

Since bandwidth doesn't seem to be of concern, LACP LAG seems to be what you're looking for. If you also use VLAN trunking, configure that on top of the LAG interface.

VRRP is something different: it creates a redundant, virtual router by using a failover scheme between multiple physical routers. That way, your end nodes can use a single uplink router without having to worry about the failover between multiple gateways themselves.

Not the answer you're looking for? Browse other questions tagged or ask your own question.