vMotion fails to communicate between certain hosts but not all
Article ID: 419950
Updated On:
Products
Issue/Introduction
vMotion communication, either through vMotion tasks or vmkping testing, works only to certain hosts in the environment. See Testing VMkernel network connectivity with the vmkping command for more information about vmkping.
The hosts that cannot be reached are able to communicate with each other normally; in other words, the network seems to be partitioned where hosts in one group can communicate between each other, but not to the other group.
There are two physical switches in the physical network that the hosts are using, and the vMotion network may be set up to be able to use at least one NIC on either physical switch.
Environment
vSphere (all versions)
Cause
The physical switches are not able to communicate between each other, at least over the vMotion network.
Resolution
Work with the physical network team and/or vendor to determine why cross-switch traffic is failing, such as with a missing VLAN configuration on a Virtual Port Channel (VPC) between the switches.
Additional Information
This behavior can be isolated by determining if the hosts that can communicate with each other are leveraging the same physical switch for the vMotion traffic. For instance, if vmnic1 on all the hosts is connected to physical switch "Switch-A", and vmnic2 is connected to "Switch-B", and the vMotion adapter is set to Active/Active with vmnic1 and vmnic2, then all the hosts using vmnic1 can communicate with each other but not with any using vmnic2 and vice-versa.
If a host that is using a specific switch begins using the other NIC, for example by bringing down the current NIC for that host's vMotion network forcing a failover to the other, it can now communicate with hosts in the new group but not the old.
NOTE: To determine the NIC a VM or vMotion adapter is using in real-time, open an SSH session to the host in question and type "esxtop" then press the "n" key to view the networking page. The in-use NIC will be in the TEAM-PNIC column to the right of the vMotion adapter, e.g vmk1.
In the below example, we can see pings over the vMotion network to the partner host initially failing, then succeeding once the host has been moved to the other physical NIC:
Feedback
