Switching

Just thinking through the issues with regard to having a 4-member VC (any EX series) and using 2, or more, of the management ethernet links connecting to a single management switch.  To illustrate:

EX4200 VC - myVC-1:

Member-1 (RE)

Member-2 (BK)

Member-3 (LC)

Member-4 (LC)

EX2200 - Management switch:

Manage-1

If I take the management ethernet ports of both "Member-1" and "Member-2" and connect them both directly to the standalone switch, "Manage-1," do I need to do anything special to prevent loops, eg. RTG or STP on "Manage-1"?  Or will this be taken care of internally by "myVC-1" in a similar way to vmware's vSwitch, in that the virtual MAC address used for the vme interface is only presented to one physical (management ethernet) interface until that interface fails?  Presumably, in that scenario, "Manage-1" would only ever see the vme MAC address on one of the ports connected to myVC-1 so it would be safe to leave STP/RTG out of the equation.

Is this correct?  Or would it always be recommended to use some form of redundant link management feature on "Manage-1" as an extra precautionary method of loop-avoidance?

Interested on your thoughts and experiences.

Cheers.

I'm doing exactly that with a 3-member VC.  No issue at all.  Think of it like a LAG.

These are the three onboard me0 ports in the VC:

Interface    Link     Input bytes        (bps)      Output bytes        (bps) Description
 ge-0/0/17     Up          152510      (11568)            170028      (14752) ex4200-fan-1.0
 ge-0/0/18     Up               0          (0)             21090       (1024) ex4200-fan-1.1
 ge-0/0/19     Up               0          (0)             21090       (1024) ex4200-fan-1.2

While the switch the VC is attached to does send stuff out (LLDP in this case), the VC itself only uses one of them.

-Chad

Many thanks for the response, Chad.

Documentation is super-unclear about this.

I've read the above answer made by Chad.

Anyway, I'd like to double check : when a VME interface is configured (for instance here in a 4 members VC configuration), and when every member as a connection between its me0 and an external switch, there is no risk of creating an Ethernet loop between the VC and the external switch?

member 0 RE(me0) ===> Mgt-SW port 1

member 1 BK(me0) ===> Mgt-SW port 2

member 2 LI(me0) ===> Mgt-SW port 3

member 3 LI(me0) ===> Mgt-SW port 4

How does it work ? When a Vme interface is configured, the master me0 interface is active and every other me0 (on backup and linecards members) is in kind of blocking mode ?

So, in case the master goes down, the backup take precedence and its me0 become active ?

Keep in mind that the me0/vme interface should be in family inet/inet6, with an ip address directly assigned to it, not in family ethernet-switching (if that is even possible).  When an interface is configured as family inet/inet6 it will directly consume/ignore incoming packets.  It is only when an interface has family ethernet-switching applied does the normal switch behavior occur, where an incoming packet is then sent out other port(s) in the same broadcast domain/vlan.

Like all of the other interfaces with family inet/inet6, the ip address is on the master RE.  For the me0/vme port, that means the master's physical me0 port will be used for the vme interface.  It -might- move to one of the other members if that link drops; I haven't tested.  I think there was a kb article that said that it is possible to use the individual member me0 interfaces by assigning an ip address to each one.  I believe it may even have mentioned it being feasible to use vme in parallel with individual me0 interfaces but it required doing something outside of the main cli.

Also, the virtual chassis links, whether using VCP cables or onboard 1G/10G interfaces, will not pass user traffic until the adjacent switches negotiate that they are part of a virtual chassis.  That means no switch loop between the members until they both agree to behave as a single switch.

Does that help?

-Chad

Does that help? => A lot ! Thank you very much !

"Family inet" means no loop, that should have been obvious :-x

For the vme address moving from a defective member (master) to another, that's how it is designed to work, insn't it ? (sorry I've got no testing environment, maybe in a few weeks, i've asked for a pair of ex4200 for testing purpose but things go slow when it comes to justify a purchase ...).