Does MLAG(SN2010) support two non-LAG connections?

[hiroyuki_sato]

Hello, members.

I have some questions about non-LAG connections in the MLAG environment.
(I don’t have switches. so I can’t send the configuration. I just want to know the specifications.)
I referenced this document.
https://enterprise-support.nvidia.com/s/article/how-to-configure-mlag-on-mellanox-switches

• Is it possible to connect between PC1 and PC2 (Both hosts connect VLAN 100, without LAG) (Use two SN2010 switches only.)
• If possible, How Switch1(PC1) sends a packet to Switch2(PC2)? Does It use an IPL port? Do I need to connect two switches like port 18?
• Does anyone know these documents and configuration examples for non-LAG connections?

Best regards.

[eddies]

• Is it possible to connect between PC1 and PC2 (Both hosts connect VLAN 100, without LAG) (Use two SN2010 switches only.)
  Yes

• If possible, How Switch1(PC1) sends a packet to Switch2(PC2)? Does It use an IPL port? Do I need to connect two switches like port 18?

The IPL is used for this communication

[hiroyuki_sato]

Hello, @eddies Thank you for your reply.

Please tell me a little more.

• Do both switches must connect each mgmt port?
• Does MLAG work correctly if each mgmt port is disconnected?
• Do I need to add VLAN to the IPL port like red text(IIUC, NO)
• Does all that have to do is set the MLAG configuration? (for non-LAG ports.)

Best regards.

[eddies]

Please tell me a little more.

• Do both switches must connect each mgmt port?
  it’s not a must - using the mgmt ports (connected to the same mgmt network) - there is the option to detect a split brain scenario
• Does MLAG work correctly if each mgmt port is disconnected?
  yes - without the split brain detection option
• Do I need to add VLAN to the IPL port like red text(IIUC, NO)
  no
• Does all that have to do is set the MLAG configuration? (for non-LAG ports.)
  yes

[hiroyuki_sato]

Hello, @eddies. Thank you for your help.

Please tell me a little more. (Probably, last questions)

• Connecting mgmt ports avoid split-brain. correct?
  • Without mgmt connection, It causes split-brain when IPL ports are down. (up image)
  • With mgmt connection It avoids split-brain when IPL ports are down(down image).
• How to detect switch1 peer is Switch2 (If IPL ports are down)? Does It use mlag-vip, IPL interface IP, or Peer mac-address?
• Does MLAG VIP use VRRP or proprietary protocol?
• In this image, Switch2 becomes master because It uses the highest IP. Correct?
• master election(below description), If switch1 becomes master, It doesn’t change to slave automatically (like VRRP priority). Correct?
• If each mgmt connection is down, but IPL ports work correctly, MLAG still works correctly. Correct?

Best regards.

https://docs.nvidia.com/networking/display/onyxv3103100/mlag#src-96897998_MLAG-UpgradingMLAGPair

Master election in MLAG is based on the IPs of the nodes taking part of the MLAG. The master elected is that which has the highest IPL VLAN interface local IP address.