Memory problems occur when a host uses more than 1016 dvPorts on a vDS

A few people have been talking about this lately, and what it actually means. Within the vSphere 4.1 U1 release notes the following statement has been made:

• Memory problems occur when a host uses more than 1016 dvPorts on a vDS * Although the maximum number of allowed dvPorts per host on vDS is 4096, memory problems can start occurring when the number of dvPorts for a host approaches 1016. When this occurs, you cannot add virtual machines or virtual adapters to the vDS. Workaround: Configure a maximum of 1016 dvPorts per host on a vDS.

So what does this mean exactly? If a portgroup is created with 2000 dvPorts, how would this be created on a host? Would it create 2000 dvPorts for every host in the vDS?  The Answer is NO.  A host will only have the maximum number of configured proxy ports, which by default is 256.  Proxy ports are used to map to one of the vDS ports when a VM is running on that host.  If you run out of proxy ports you will not be able to add more VMs to the vDS on that host.

A good point to note is that if you run out of vDS Proxy Ports then you are likely to get this kind of error “Virtual NIC with hardware address 00:1D:09:09:9B:E9 failed to connect to distributed virtual port 2 on switch 74 65 73 74 00 00 00 00-00 00 00 00 00 00 00 01. No more ports available on the host proxy switch.” This KB article provides a bit more in depth content how to resolve this issue.

It has been mentioned that you might be able to change the Proxy Port default option through the UI, but I have yet been unable to find this setting.  If I do I will update the article.