Consider the scenario where TrueNAS Scale is connected to 2 different switches via a Multi-Chassis Link Aggregation Group (aka MCLAG, MLAG, ICCP, etc). I have just this scenario where I have two Top-of-Rack switches running SONiC with ICCPD enabled. I am trying to use 4 ports on the TrueNAS with 2 ports connected to each ToR switch. The switches are setup for MCLAG with TrueNAS seeing just a single LACP LAG accross all 4 ports. TrueNAS and the ToR switches reside at a CoLo Data Center and the ToR switches provide L3 routing back our headquarters network. When I configured TrueNAS, I had to set the default gateway as the IP address for Switch1. But, if Switch1 ever goes down, the LAG to TrueNAS would stay up, but TrueNAS cannot connect to the default gateway because it is down.
In Windows Server, there is a client-level solution because it is possible to configure more than one default gateway. Windows Server will automatically failover to the next default gateway if the primary ever goes down. This is perfect for my CoLo network. But, on TrueNAS, I'm looking for the solution to provide the same redundancy.
Some enterprise switches (I believe Cisco and Juniper) have an IP Address sharing/MAC Synchronization feature where the MLAG switches share the same IP address. When one switch goes down, the IP address is handled/assumed by the remaining switch. This provides the default gateway redundancy at the switch-level instead of at the client-level. I have a feeler into the SONiC community to see if there is a similar switch-level feature to handle the MLAG default gateway redundancy.
Does anyone have guidance on how to achieve client-level default gateway redundancy when using MLAG to TrueNAS Scale?
In Windows Server, there is a client-level solution because it is possible to configure more than one default gateway. Windows Server will automatically failover to the next default gateway if the primary ever goes down. This is perfect for my CoLo network. But, on TrueNAS, I'm looking for the solution to provide the same redundancy.
Some enterprise switches (I believe Cisco and Juniper) have an IP Address sharing/MAC Synchronization feature where the MLAG switches share the same IP address. When one switch goes down, the IP address is handled/assumed by the remaining switch. This provides the default gateway redundancy at the switch-level instead of at the client-level. I have a feeler into the SONiC community to see if there is a similar switch-level feature to handle the MLAG default gateway redundancy.
Does anyone have guidance on how to achieve client-level default gateway redundancy when using MLAG to TrueNAS Scale?
