4 way smlt / rsmlt - routing redundancy

[BRMS]

I’m migrating our backbone to 4 vsp 8600 switches; two per location. for servers we are using a single layer 2 vlan that is spanned over both locations. for vlans that only exist in one location i use an rsmlt configration on the two corresponding vsp-switches to make sure there is no single point of failure.

i would like to create the same configuration for the server subnet, so that every vsp can act as their router. is it possible to create a 4 way rsmlt cluster between the 4 vsp for the server vlan? do i need to use vrrp for that? maybe a combination of both? whats the best practice for the above configuration?

[Douglas12]

It will take significant time for the failover switch to get all multicast routes populated. Before both unicast and multicast protocols get convergent, traffic loss will be seen.

tellthebell

[BRMS]

you are exactly right about my configuration. the solution sounds nice and i already tried it; seems to work. i’m wondering if i should only use vrrp for that vlan or if vrrp and rsmlt are both needed. the configuration im currently testing is:

    parents: "interface Vlan {{ vlan_id|int }}"
    lines:
      - "ip address {{ vlan_subnet }}.{{ server_subnet_gw_ip }} {{ vlan_netmask }}"
      - "ip rsmlt"
      - "ip vrrp version 3"
      - "ip vrrp address {{ vlan_vrrp_id }} {{ vlan_vrrp_ip }}"
      - "ip vrrp {{ vlan_vrrp_id }} backup-master enable"
      - "ip vrrp {{ vlan_vrrp_id }} priority 200"
      - "ip vrrp {{ vlan_vrrp_id }} enable"

i hope the yaml-notation in my configuration example is not a problem, i copied it out of my ansible role. every device gets a unique ipaddress, the vrrp_id and vlan_vrrp_ip are identical on all four.

im not entirly sure if rsmlt is still needed on that vlan if i already use vrrp.

another question: how to distribute priorities between the vrrp nodes? one per location with 200 and one per location with default (100)? all four the same? 100,200,300,400?
 

this article: https://gtacknowledge.extremenetworks.com/articles/How_To/How-to-Configure-VRRP-with-BackupMaster-on... mentions i need to enable pinging of the virtual address with ping-virtual-address enable. i didn’t do that but still can ping the ip, how can that be?

when executing “show running-config” the vrrp global part is empty, which seems rather odd:

------------------------------------------------------------------------

#

# VRRP CONFIGURATION - GlobalRouter
#

#
# VRRP CONFIGURATION - VRF
#

#
# UDP FORWARDING CONFIGURATION - GlobalRouter
#

________________________________________________________

[EXTR_Paul]

I assume your four VSP8600s are deployed as two vIST Clusters? Interconnected with SPBm NNIs?

You are correct.  RSMLT is only support within two switches in an IST/vIST cluster.  

So its ideal for a two switch deployment you would see an in a small core or distribution cluster.

In your case, to have all four VSP8600 to share redundancy for that server  vlan you MUST you use VRRP.  

The sweet thing is that RSMLT and VRRP are interoperable.   you will have no issues deploying this. 

Here is a best practices guide.  

https://documentation.extremenetworks.com/TCG-TSG/NN48500-617_ShortPathBridg802.1aq__SRG_Dec2017.pdf