PCoIP works on one subnet but only RDP works on another?

[Newbie009]

Hi all,

I don´t know why this is happening but if anyone can help I would appreciate it.

My set-up:

Subnet A has: ESX Server, VCenter Server, VMview Connection Server (with 2 NICs, NIC1 connected to subnet A, NIC2 connected to subnet B - there is no communication between subnets A & B), template for subnet A clients, DHCP Server

Subnet B has: DHCP Server, template for subnet B clients (with host file entry for NIC 2 on VMview Connection Server).

Created 2 pools, PoolA uses a template with a NIC connected to subnet A, PoolB uses a template with a NIC connected to subnet B.

Test user has permissions to use both pools.

Using test user on a physical client machine in subnet A:

Connecting to PoolA, I can connect to VMview clients with PCoIP fine.

Connecting to PoolB, when I connect using PCoIP it appears to connect but then "hangs" at a black screen. I can minimize the black screen and maximize again but it doesn´t change. Also the screen does not disappear after 5 seconds (as I have read in other set-ups).

If I however change the protocol to RDP, I can connect to PoolB fine and without problems.

I tried looking at an article that was recommended on these forums with a list of things to check if you see the black screen but everything appears to be setup correctly.

Does anyone have any ideas as to what could be causing this?

Is this (as I think) something related to the fact that the subnets don´t have communication between them?

I haven´t tried yet as I am off-site again but do you think this behaviour would not occur for PoolB if I was connecting with a physical client directly connected to subnet B (using the IP address for the connection server of NIC2 on the Connection Server)?

Any ideas?

Thanks

Mark

[rosspcs]

How are you allowing the communication between subnet A and B - is it a firewall, if so i would start there and amke sure the firewall is allowing UDP through

[Newbie009]

Hello Rosspc,

Thanks for the reply.

As far as I am aware, the client here has 2 completely isolated subnets i.e. no communication between them at all.

What I did to configure things was to add a second NIC to the VMview Connection Server and connect this by cable directly to subnet B.

Then when I created a plantilla for a machine in subnet B, I connected it´s NIC to subnet B and created a host file entry for the Connection Server of the IP for the second NIC on the Connection Server.

Basically they have 2 unconnected subnets and they want to deploy VMview clients to both.

Mark

[rosspcs · Accepted solution]

So you are using the connection server as a bridge, which NIC has the default gateway ? i am think that you may need to add some static routes on the connection server. it is most likley that the problem is with UDP thats would explain why RDP works.

[Newbie009]

Hello Rosspc,

Thanks for the reply.

As far as I am aware, the client here has 2 completely isolated subnets i.e. no communication between them at all.

What I did to configure things was to add a second NIC to the VMview Connection Server and connect this by cable directly to subnet B.

Then when I created a plantilla for a machine in subnet B, I connected it´s NIC to subnet B and created a host file entry for the Connection Server of the IP for the second NIC on the Connection Server.

Basically they have 2 unconnected subnets and they want to deploy VMview clients to both.

Mark

[Newbie009]

Hello Rosspc,

Thanks for the reply.

As far as I am aware, the client here has 2 completely isolated subnets i.e. no communication between them at all.

What I did to configure things was to add a second NIC to the VMview Connection Server and connect this by cable directly to subnet B.

Then when I created a plantilla for a machine in subnet B, I connected it´s NIC to subnet B and created a host file entry for the Connection Server of the IP for the second NIC on the Connection Server.

Basically they have 2 unconnected subnets and they want to deploy VMview clients to both.

Mark

[Newbie009]

Hello Rosspc,

Sorry about those other repeated posts - problems with my browser....

I have both NICs configured with a gateway - NIC on subnet A is configured with GW for subnet A and NIC for subnet B is configured with GW for subnet B.

Which direction do I need to put the route? For traffic with a source of subnet A and a destination of subnet B?

If I understand correctly then, the connection server needs to know that it should send packets for subnet B to the second NIC. I guess they are by default being sent to the GW on subnet ?

I will look into creating a persistent route for this.

I will let you know how I go.

Cheers

[rosspcs]

So one thing you dont want is 2 default gateways that will cause problems, i would take the gateway off of NIC B.

[Newbie009]

Hi again,

OK so I removed the default GW from NIC 2.

I have attached a doc with a screenshot of the IP settings for NIC2 and the latest route print output.

I just tried another test for the client and it is still showing just a continuous black screen.

Is my routing set up correctly?

Thank you for helping.

Mark

[rosspcs]

The static route should have the default gateway of subnet B not NIC B

[Newbie009]

Ah, Doh!

I am not onsite anymore so I will correct this when I return and try again.

Thanks again for your time.

[Newbie009]

Hi Rosspcs,

I have not been able to return to the client´s site yet and it might be a while but your suggestions seem to ascertain the cause.

Thank you for your help on this.

Regards

Mark

PS: I have awarded the points to you as you helped a lot.

[mpulsfort]

There may be another thing to be aware off.

When using RDP you are connecting with the clients through the connection server and the connections are tunneld through the connection server.

When using PCoIP the Clients in Subnet B need a direct UDP connction to the VMs running on the ESX Hosts.

As I understand your posts, thats not the case in your enviroment.

[Newbie009]

Hello mpulsfort,

Thank you for your response.

So what you are saying is that basically any deployed VMview client that will use PCoIP requires a direct UDP connection to the ESX Host that is runing the virtual desktop mchine. Correct?

In that case the site I am trying to configure, unless they implement some sort of communication between the 2 subnets, will only be able to use RDP.

Is the only way to get the UDP access through a router (hardware or software) or is there another way?

Please correct me if I misunderstand and thanks again.

Mark

[mpulsfort]

Yes that`s correct.

As I understand your virtual Desktops are Running in Subnet A. So if you try to connect from Subnet B to one of your virtual desktops via PCoIP

you need a direct UDP Connection to that desktop.

You can place the virtual dekstops in Subnet B which may not be acceptable for you or you have to establish some sort of communication between the subnets.

[Newbie009]

Ok, thanks very much.

Mark

[rosspcs]

That's not strictly true firewalls and routers will pass udp and a multihomed server can if its configured correctly, is bridging enabled and is routing and remote access installed

[Newbie009]

Hi guys,

I was using the VMview Connection Server with 2 NICs to connect the subnets but RRAS was not enabled. I was not aware of exactly how the VMview clients communicated with the Servers and, in this environment, it caused some confusion.

My tests have now actually finished as we collected all the equipment yesterday without allowing me to try to rectify this issue.

But the client was more than happy with PCoIP for the main subnet and just RDP for the secondary subnet.

I gave them some recommendations and tips for any possible future implementations there.

Thanks for the help from both of you.

Regards

Mark

[ArniZLV]

Hello!

We have similar problem - Virtual Desktop machine has two network connections, one for internal network, one dedicated for external server connection. And the problem is that PCoIP listens on connection that is on external network, which is not accessible from other machines in internal network. So VMView client can not connect to PCoIP server. When that VD vas booted only with internal NIC everything was working fine. Is there any way to tell the ViewAgents PCoIP server to use internal NIC?

[markbenson]

All this can be solved using the PCoIP Secure Gateway feature which came in with View 4.6.

It is designed to bridge different network address spaces. It is primarily designed for remote access to a View environment from the Internet where there is a different address space on the Internet to the address space in your DMZ/Internal network. This works in the same way on an internal network where you need to route clients on one network to virtual desktops on another.

View works in this way for both PCoIP and RDP and so you can set up a View environment that allows access from both networks and supports RDP and PCoIP as needed.

Look at this document http://communities.vmware.com/docs/DOC-14974 and look at the video which details how to set it up. This describes access from the Internet, but it is just the same for two internal networks with different address space (different subnets). The worked example (about 18 mins in) shows the setup to support access from both networks to the same view environment (internal and external).

If, after watching the video, you still have trouble configuring it, this forum will help.

Mark.