Question

Question about network high availability

  • 27 July 2022
  • 1 reply
  • 24 views

Hello, 

Case 2 is not working normally in the AHV cluster
Is there a solution?

1. Cluster infomation
 > AOS : 5.20.4.5
 > Cluster Network : br0(vs0)
 > Service Network : br1(vs1)

2. HA test scenario
 2-1 Case 1 : AHV01 br0 remove two cables 
  > Result : Virtual machine move to another host
 2-2 Case 2 : AHV01 br1 remove two cables 
  > Result : Virtual machine does not move to another host

thanks


1 reply

Userlevel 4
Badge +5

hi @Jae Ho 

In case 1 is the expected behaviour. CVM’s eth0 is connected to br0/vs0 and if you unplug both cables from the host, it fails (because of the loss of network connection) and system starts self healing re-creating another data copy to recover resiliency and it restarts UVM’s on another host.

In case 2 you are using br1 for a set of VLANs for the UVMs. When you disconnect the cables CVM of that host is working and it has normal network connectivity because it goes through vs0 so for that reason the host doesn’t “fail” and UVMs have no network connectivity. I’ve never run on that scenario on a production environment but, maybe you start receiving alerts from ncc because of UVMs pinned to a VLAN on a disconnected vs.

Here’s and example. I have lab hosts with vs0 (purple) and vs1 (green) and normally vs1 cables are disconnected (only used for a few tests) cluster works perfectly but if I try to pin a UVM to a VLAN created on vs1 obviously it won’t have network connection

To sum up. You have at least a pair of cables (Base-T, TwinAX...) connected to each VS. You need to ensure that at least one switch or one cable is available ALWAYS!

Hope this clarify things a bit

Regards!

Reply