Hi All,
We've purchased a Dell PowerEdge 1950 and we've been trying for couple of weeks to run 2 Windows 2003 guest OSs. The virtual machines run fine for an undefined period of time and then connectivity is lost (the guest OSes do not respond to pinging), the host OS is still fine. After about 10-15 minutes the guest machines come back to life as nothing ever happened and they did not crash, there is absolutely nothing in the event logs. The guest servers use "bridged networking" and use dynamic IPs, the networks cards in the server are Broadcom BCM5708C NeteXtreme II, and the we use VMware Server 1.0.4. We've tried various versions of the host OS (Windows 2003 32-bit Standard and Enterprise editions and also Windows 64-bit Standard and Enterprise editions) and various versions of the network drivers from Dell and Broadcom. Same issue.
At the same time we were researching the issue on the internet, it seemed that some people point out that this behaviour might be caused by the new TOE (TCP/IP Offload Engine) that's in the Broadcom NICs. So we followed the procedure below supplied by Dell. Same issue.
Now, the identical 2 guest machines work just fine on two different machines and they also run fine on one other machine. So far one common thing that we have noticed when researching this issue is the use of the Broadcom NeteXtreme II NICs. I would say to stay away from those network cards if you're purchasing a new server and using it with VMware. I am tempted to try Microsoft Virtual Server just to narrow down the problem, however due to time constraints and the return policy on the Dell server I will not have time to try it.
At this point, I think I'll return the server and do some more research before buying a new one.
Please let me know if anyone has resolved this issue using a "real" solution, not half-measures (i.e. constant pinging of the guest machines or switching to NAT for VMware network configuration)?
Thanks,
Arthur
======================================================================
For the interested, here is the Dell procedure to disable TOE:
How to disable TOE
1) Uninstall the Broadcom driver from the add-remove programs and power down the server,
2) Remove the TOE Key by following these steps.
Removing the TOE Key
Before removing the TOE key:
a. Disconnect the system from AC power and Remove the top cover.
b. Locate the TOE key.
<<DSC05058[1].jpg>>
c. Press in on the TOE key RJ-11 release latch.
<<DSC05060[1].jpg>>
d. Lift the TOE key out of the chassis.
3) Power on the server and reinstall the latest Broadcom driver.
4) Follow the steps of the article : http://support.microsoft.com/kb/927695
5) Follow the steps of the article http://support.microsoft.com/kb/904946
6) Turn off the TCP Chimney
To turn off TCP Chimney by using the Netsh.exe tool, follow these steps:
a. Click Start, click Run, type cmd, and then click OK.
b. At the command prompt, type Netsh int ip set chimney DISABLED, and then press ENTER