This article provides information about multihoming VMkernel networking.
Multihoming in a VMkernel networking context means that there are multiple VMkernel adapters (vmknic) in a single TCP/IP stack (Netstacks).
Observed issues that may be present:
- vMotion operations can fail or take an exceedingly long time.
- Storage latency may be reported (iSCSI not using port binding).
- ESXi host disconnects from vCenter can occur.
- ICMP replies to vmks may not be received.
- vSphere/vSAN Health Check failures, including "vMotion unicast connectivity check", "vMotion MTU check", or "vSAN MTU check" alarms.
The above may exist if you have VMkernel ports configured on the same IP subnet and TCP/IP stack similar to:
- Managment vmk0 with IP 192.#.#.8/24 and Default Gateway 192.#.#.1 as part of the default TCP/IP stack.
- vMotion vmk1 with IP 192.#.#.10/24 and Default Gateway 192.#.#.1 as part of the default TCP/IP stack.
- iSCSI vmk2 with IP 192.#.#.12/24 and Default Gateway 192.#.#.1 as part of the default TCP/IP stack.
In the above configuration, all of the vmknics are in the same IP subnet and TCP/IP stack. This results in the TCP/IP stack potentially selecting the wrong vmknic for egress vmknic traffic and can lead to potential issues with connectivity, low throughput, and asymmetric routing