Change ISC-link between MLAG peers


Userlevel 3
Hi everybody.

I'd like to change ISC link between two MLAG peers for some reason. It would be great if someone gave me a few tips about order of work.

I have two x680 switches and LAG between them. It contains four 40G links (VIM4-40G4X-1). This is my ISC link. I'd like to change this LAG to two 10G ports.

I'm a bit confused about it. Especially about order of work.

Should I follow this plan:
1. shutdown four 40G ports
2. delete mlag peer and all other MLAG configuration
3. create new LAG (2x10G). It'll be the future ISC link.
4. create new MLAG peer
5. enable necessary MLAG ports

My simple MLAG configurationcreate mlag peer "xCore2"
configure mlag peer "xCore2" ipaddress 10.255.255.254 vr VR-Default
enable mlag port 1 peer "xCore2" id 1
enable mlag port 2 peer "xCore2" id 2
enable mlag port 3 peer "xCore2" id 3
enable mlag port 4 peer "xCore2" id 4
enable mlag port 5 peer "xCore2" id 5
enable mlag port 6 peer "xCore2" id 6
enable mlag port 7 peer "xCore2" id 7
enable mlag port 8 peer "xCore2" id 8
enable mlag port 36 peer "xCore2" id 107
enable mlag port 37 peer "xCore2" id 106
enable mlag port 39 peer "xCore2" id 105
enable mlag port 40 peer "xCore2" id 103
enable mlag port 42 peer "xCore2" id 104
enable mlag port 44 peer "xCore2" id 102
enable mlag port 46 peer "xCore2" id 101 [/code]

8 replies

Userlevel 7
You create your new LAG (2x10G) and you add it to the ISC VLAN (without cabling the ports) and you configure the necessary vlans on it too.

The idea will be to unplug the 40G links and plug the 10G links. But for some time, you have no ISC.

If everything is dual-homed :
> If you are running 15.5+, you can configure the alternate health-check path feature to handle automatically this situation.
> or you shut the MLAG port yourself on one MLAG peer.

You remove the 40G links, then plug the 10G links. You remove the 40G LAG from the ISC vlan.

And alternate way to do it, that is not validated by our SQA and *should* work, is to use SRP feature. Once again you create a new LAG for your 10G links, add it to ISC VLAN and add the relevant vlans to it. You configure this port as the backup port fo the 40G LAG in SRP (Software Redundant Port). You plug the 10G links. Once it's ok, you unplug the 40G links.
Userlevel 3
Grosjean, Stephane wrote:

You create your new LAG (2x10G) and you add it to the ISC VLAN (without cabling the ports) and you configure the necessary vlans on it too.

The idea will be to unplug the 40G links and plug the 10G links. But for some time, you have no ISC.

If everything is dual-homed :
> If you are running 15.5+, you can configure the alternate health-check path feature to handle automatically this situation.
> or you shut the MLAG port yourself on one MLAG peer.

You remove the 40G links, then plug the 10G links. You remove the 40G LAG from the ISC vlan.

And alternate way to do it, that is not validated by our SQA and *should* work, is to use SRP feature. Once again you create a new LAG for your 10G links, add it to ISC VLAN and add the relevant vlans to it. You configure this port as the backup port fo the 40G LAG in SRP (Software Redundant Port). You plug the 10G links. Once it's ok, you unplug the 40G links.

I'm sorry but what does "SQA" meen?
Userlevel 7
Grosjean, Stephane wrote:

You create your new LAG (2x10G) and you add it to the ISC VLAN (without cabling the ports) and you configure the necessary vlans on it too.

The idea will be to unplug the 40G links and plug the 10G links. But for some time, you have no ISC.

If everything is dual-homed :
> If you are running 15.5+, you can configure the alternate health-check path feature to handle automatically this situation.
> or you shut the MLAG port yourself on one MLAG peer.

You remove the 40G links, then plug the 10G links. You remove the 40G LAG from the ISC vlan.

And alternate way to do it, that is not validated by our SQA and *should* work, is to use SRP feature. Once again you create a new LAG for your 10G links, add it to ISC VLAN and add the relevant vlans to it. You configure this port as the backup port fo the 40G LAG in SRP (Software Redundant Port). You plug the 10G links. Once it's ok, you unplug the 40G links.

That's our team validating that everything works fine, by testing a lot of situations and protocols combinations.
Userlevel 3
Grosjean, Stephane wrote:

You create your new LAG (2x10G) and you add it to the ISC VLAN (without cabling the ports) and you configure the necessary vlans on it too.

The idea will be to unplug the 40G links and plug the 10G links. But for some time, you have no ISC.

If everything is dual-homed :
> If you are running 15.5+, you can configure the alternate health-check path feature to handle automatically this situation.
> or you shut the MLAG port yourself on one MLAG peer.

You remove the 40G links, then plug the 10G links. You remove the 40G LAG from the ISC vlan.

And alternate way to do it, that is not validated by our SQA and *should* work, is to use SRP feature. Once again you create a new LAG for your 10G links, add it to ISC VLAN and add the relevant vlans to it. You configure this port as the backup port fo the 40G LAG in SRP (Software Redundant Port). You plug the 10G links. Once it's ok, you unplug the 40G links.

Thanks. I got it.
Userlevel 6
great suggestions Stephane. Eyev are you planning to do this in production or during a shutdown window. I would recommend a window although things should not be a problem doing what Stephane has recommended you could get yourself into a loop condition and have a big mess.

I would also create two scripts one to set up your LAG and ISC ports and one to remove them. That way it is fast can be done on both switches at the same time and can be easily undone if you need to.

Just being paranoid 🙂

P
Userlevel 3
Thanks guys! Sounds very simple.

Create new LAG, add it to ISC VLAN, add necessary VLANs to it, shutdown old ISC LAG link, plug new ISC link. I'm going to plan a maintain window!

Anyway, I will explain why I have to do it. We have two x670 in core and we've decided to move one of them to another geographical location. Redundancy, you know. But we don't have any QSFP+ 40G LR modules unfortunately. Thats why we should create 2x10G LAG between the core switches.
It's not a problem to shutdown one x670 switch and reconfigure MLAG, because we have to deliver it to a remote location.

Sorry for my English...
Userlevel 7
Just to state the obvious, if SRP is not activate, don't have the 2 LAG plugged at the same time : that would cause a loop.
Userlevel 3
Grosjean, Stephane wrote:

Just to state the obvious, if SRP is not activate, don't have the 2 LAG plugged at the same time : that would cause a loop.

Yes, I see. Thank you again.

Reply