Management
Management

MCLAG between two virtual chassis mixed EX series

[ Edited ]
‎12-20-2017 02:17 AM

Hi,

 

I have the following network topology:

I have 6 network elements, two EX4600 and four EX4300, all these elements make up two independent mixed virtual chassis each formed by: an EX4600 master and two EX4300 line cards, I am trying to connect these two independent VCs to each other through MCLAG.

Is this possible in the EX series? If not, can you tell me how to put them in HA without being a master / slave topology?

Regarding my configuration, I have adapted the following link to my network: https://www.juniper.net/documentation/en_US/junos/topics/example/multichassis-link-aggregation-qfx-s...

In my case I configure only VLAN 50 without levels 3 and connected an ESXi with a cable to each 4600 forming a LACP.

In one of the VC, let's call it, VCA, I have connected, an ESXi(ae1) cable, a HUAWEI(ae2) cabin in LACP. On the other VC, let's call it VCB, I have connected the other ESXi cable(ae1).

In each of the VCs I have two ports forming LACP for the ICCP protocol(ae0).

 

From VCB I execute the following command obtaining the following results:

 

root@EX4600-1-VC2> show ethernet-switching table

MAC flags (S - static MAC, D - dynamic MAC, L - locally learned, P - Persistent static, C - Control MAC
SE - statistics enabled, NM - non configured MAC, R - remote PE MAC, O - ovsdb MAC)


Ethernet switching table : 3 entries, 3 learned
Routing instance : default-switch
Vlan MAC MAC Age Logical NH RTR
name address flags interface Index ID
VLAN50 00:50:56:xx:xx:xx DLR - ae1.0 0 0
VLAN50 a8:ca:7b:xx:xx:xx DR - ae0.0 0 0
VLAN50 a8:ca:7b:xx:xx:xx DR - ae0.0 0 0

 

That is, the VCB is seeing what is connected in VCA through the ICCP (ae0).

I do the following test, I disconnect the ESX cable that is in the VCA to see if from the VCB, in which I only have the ESXi, I can reach the VCA, the cabin and I do not get it, why can it be?

I hope to have explained correctly, if you need some more or greater depth in the explanations tell me.

 

Regards!

4 REPLIES 4
Management

Re: MCLAG between two virtual chassis mixed EX series

‎12-20-2017 12:22 PM

Hi,

 

Can you please share your existing topology diagram.

Management

Re: MCLAG between two virtual chassis mixed EX series

‎05-17-2018 11:52 AM

I ran into same issue, with QFX5100 mixed mode VC with EX4300.

 

Then I figured, in Junos you can implement HA by Six Methods;

  • Standalone

  • Virtual Chassis Fabric (VCF)

  • QFabric Node

  • Virtual Chassis

  • Multi-Chassis Link Aggregation (MC-LAG)

  • Clos Fabric

Vitual chassis in itself is a HA method, with master and backup RE in VC. You can't map MC-LAG and Virtual Chassis methods in one design. They are two independent methods in themselves for HA. Either you can have a MC-Lag or VC design not both.

 

For more: 

https://www.safaribooksonline.com/library/view/juniper-qfx5100-series/9781491949566/ch04.html

 

 

 

Highlighted
Management

Re: MCLAG between two virtual chassis mixed EX series

‎05-17-2018 01:33 PM

You're correct on MC-LAG & VC being mutually exclusive. It will appear to work until you try to add a new MC-LAG ae interface using a non-member-0 port. We found out the hard way.

Management

Re: MCLAG between two virtual chassis mixed EX series

‎07-18-2019 08:28 AM
Obviously, it is not entirely true that VC and MCLAG are mutually exclusive features: https://www.juniper.net/documentation/en_US/junos/topics/concept/virtual-chassis-mx-series-mc-lag.ht... However, as already pointed out, it is not (yet - pleeease Smiley Happy ) a supported setup with EX/QFX VCs. I did not have the possibility to test it yet, but can hopefully do soon with QFX5100 ((1x48S+1x48T) x2). @smicker: I'd really appreciate if you could share some more detail about your experiences with VC-MCLAG. My need with the above setup actually is to only have the mc-ae interfaces on member0, which is also the vc-master. The 48t switches are solely there for connecting cu based devices; and I had some headache already with mixed VCs, so that is avoided here. That way even running the VC with only one master would not be a bad option.