Virtual Machine Goes Unreachable after vMotion or reboot

[karthikyn7]

Hi Folks,

We are running newly configured VMware ESXi 5 node cluster. Each server having 6 NICs in which 2 NICs dedicated for Management and vMotion. Rest connnected to trunk ports for multi VLAN communication. In vSwitch 2 we have configured one port group for each VLAN with total 5 VLANs.

We observed our Virtual machines becomes unreachable from network if;

-DRS vMotion the VM

-Manual vMotion

-VM rebooted

Also this occurs randomly not on any particular machine and this is intermedient problem which doesn't not occurs all time but most time.

Please help me understand this issue and for the resolution of the same. Also if this is physical network issue how can I justify it?

Thank you,

Karthi.

[rickardnobel]

From the initial amount of information I would guess the issue is incorrect VLAN configuration at your physical switches.

When you do a VM reboot or vMotion the VM is assigned a new logical connection to an outgoing physical interface (vmnic). If not all physical ports are configured the same way and allows all tagged VLANs then you will get this kind of issue.

What kind of physical switches do you have?

[sharninder]

Difficult to really answer your question without looking at the configuration. From your description, everything seems fine but I wouldn't make a guess. Since you've also mentioned that VMs become unreachable on a reboot sometimes, I think it may be something as simple as an IP address conflict? In anycase, this seems like a network problem and it's going to be difficult to debug it without more info.

[a_p_]

Some more things to double-check:

• ensure the physical ports do not have port security enabled (the default for some switches)
• ensure spanning-tree is configured properly for the uplink ports (e.g. spanning-tree portfast for Cisco, Rapid Spanning-Tree for other vendors)

André

[rickardnobel]

ensure spanning-tree is configured properly for the uplink ports (e.g. spanning-tree portfast for Cisco, Rapid Spanning-Tree for other vendors)

A smaller detail, but might be valuable: Rapid Spanning Tree is not really equalient to Cisco "portfast". RSTP is a newer version of the entire Spanning Tree protocol, which has both quicker total calculations in network topology changes (few seconds compared to up to 50 seconds on original STP) and very fast client port startup into forwarding state.

For a switch operating in RSTP mode you can if needed put a port into a hard "edge" mode, quite similar to "portfast", however often not really needed due to the faster overall workings of RSTP.

It is certainly very recommended to make sure that "portfast" is used in original STP switches, like Cisco default, however it should also only matter when the link physically goes up or down. In cases like VM starts / vMotion and similiar the physical link is unchanged and will not make either STP or RSTP aware of any need for action.

[a_p_]

... but might be valuable.

Absolutely valuable and highly appreciated. Thanks for the details.

André