Vmotion and network connectivity

[Sangokan]

Hi,

I have a strange thing happening. I have vmotioned some VMs just now (around six machines) and two of them have lost connectivity (impossible to ping any VMs or workstations from those to vmotinned VMs), the only fix i have found now is to reboot the VMs...Any ideas what is wrong?

Edit: as a matter of fact this is only happening on a 3.0.2 esx host, is there's a bug somewhere? My other hosts are still in 3.0.1 and i havent seen this happening yet.

Message was edited by:

Sangokan

[oreeh]

Do you have "notify Switches" enabled in the vSwitch configuration?

[Sangokan]

Yep, on all switches

[mmcdermott]

I am seeing the same issue on 3.0.2 regarding networking. I have the notify switches enables as well.

[Texiwill]

Hello,

It sounds like the arp cache is incorrectly set in one of your switches. Are you having trouble pinging VM to VM on the same vSwitch or VM to VM across a pSwitch?

Best regards,

Edward

[oreeh]

Since there's no problem with ESX 3.0.1 hosts I doubt that it is MAC/ARP cache related.

[mmcdermott]

Thats a good question. I will have to check that next time it happens if i have connectivity within the vswitch. I am in the testing phases of our environment, so i currently have all my vms pulling a dhcp address since they are not production yet. From the VM network to the rest of the lan is certainly an issue since the vms won't pull an address during that time.

Hello,

It sounds like the arp cache is incorrectly set in

one of your switches. Are you having trouble pinging

VM to VM on the same vSwitch or VM to VM across a

pSwitch?

Best regards,

Edward

[Sangokan]

The problem is not on a switch level but on a VM level, some VMs on the same switch are pinging ok other VMs while one or two are totally unable to ping any other hosts (VMs or Physical) until i reboot them after a vmotion.

For the moment it is only noticeable on VMs that have been vmotionned to my only ESX 3.0.2 hosts while other ESX are still in version 3.0.1

There is something not standard on every switches , if i look the properties of the network adapters in the Configuration tab i see different networks...we are working with 10.1 subnets with a 16 bits mask, the properties show either networks 10.1.1.1-10.1.1.63 or 10.1.1.1-10.1.1.31

Message was edited by:

Sangokan

[oreeh]

There is something not standard on every switches,...

Simply ignore this. The network detection code doesn't detect networks reliable and it only is an information anyway (all vSwitches are Layer 2 and not Layer 3).

[Sangokan]

Problem was due to hard switch config.

[TCronin]

What was the actual problem/fix in case someone else has the same issue?

[LandscapeMode]

I bet one of the vmotion port on one of the esx hosts was connected to a switch port with wrong vlan configuration (VLAN ID was wrong)

Message was edited by:

LandscapeMode

[TCronin]

That wouldn't break the guests ability to ping, the vmotion would fail.

[Sangokan]

Actually i am still having network issues. It happens that some guests loose their network connectivity after a reboot (a guest reboot). I have checked the arp settings on our pSwitches:

show mac-address-table aging-time

Vlan Aging Time

\-- \--

---

1 300

2 300

3 300

98 300

99 300

4 300

Should i change the value, is there anything else i should try?

[maia06]

Hi,

We have the same problem on ESX 3.0.1 : after vmotion some (and not all) VMs cannot be pinged from outside but it can be pinged from another VM on the same ESX server.

[bstratton]

I also have this issue where some guests on a switch lose network connectivity after a reboot or after importing.

[b18b]

Hi,

We have the same issue for both Vmotion and server reboots where the virtual machine loses network connectivity.

The issue occurs for various Operating system versions including Window 2000 Server SP4, Windows 2003 Standard SP1 and Windows 2003 Standard SP1 with R2.

ESX Version: 3.0.2

VI Client: 2.0.2

2 x ESX Hosts: IBM x3850

I have found an article in the KB for a known issue, but only for previous versions of ESX server:

Is there a resolution for this issue with ESX Version 3.0.2?

Thanks

[BDorroh]

Has anyone found a difnitive answer for this yet? I'm having the same issue. I upgraded four ESX hosts from 3.01 to 3.02 (without update 1.) When I do a warm migration from one host to another, "some" of the Vms lose connectivity. I can see this by running a ping in the console window when moving them. Moving back the other direction works fine. My fix is to just disable/enable the network adapater in the OS and that seems to fix it.