e run Workstation 9.0.1 with a set of UNIX guests matching a "standard" build image on our development notebooks and two remote desk side workstations. The major differences other than the hardware obviously are that the notebooks run on top of Windows 7 with the NAT virtual device and the two desk side systems run on top of RedHat 6.2 with a bridged virtual device. We are using bridged on the desk sides because we have one application that cannot use NAT and we use NAT on the notebooks because our VPN only allows one IP through when we work remotely.
What we are seeing on the desk side systems is that when there are multiple guests running, any network operation performed on more than two guests fails. I can go to any guest, and the first two will work, the third or greater will fail. If I recycle the interface on a filing one, it will generally start working immediately, but then one of the previous guests that was working will stop. If I let the guest all sit for a time (maybe 5 or 10 minutes, have not really timed it) , the first two I hit (doesn't seem to matter if it's RH, SuSE, Solaris, what have you) will work and the third or greater will not. We first noticed this with DNS, but soon determined that the same was true for ping or traceroute towards the nameservers.
The notebooks never display this behavior, so I'm thinking it is something to do with the bridging.
I set up a tcpdump on the host and each guest. All looks normal when it works, but on a failing guest I don't see any traffic out of the guest or host.
I've chipped away at this for a day before coming here. Been through the system logs on the host, guests and vmware logs. Piddled around with kernel tcp settings on the host. No luck so far. If I'm doing something wrong with configuration I'd rather know that now than just go for the 9.0.2 update, in the hope that something is fixed there.