Blackmagic Forum

Hi All,

at the moment we have Render Nodes 7.7.1 working on Virtual Windows 2008 Servers. Although rendering with these virtual servers is not very quick it does work. Today I tried installing Render node 8.0 on one of these Servers. The render node installed without any problems but it does not start. I see the FusionRenderNode.exe process starting but it disappears after a couple of seconds.

Is there a log file I can look into to see what the problem is?

I also tried to start the FusionConsoleNode in listen mode. Here I see the following

Code: Select all

...
Initialising GLTools
ERROR: GLTools failed to initialise!
Initialising Texture Manager

then it stops

With FusionConsoleNode 7.7.1 I also get this error but it continues

Code: Select all

Initialising GLTools
ERROR: GLTools failed to initialise!
Initialising Texture Manager
ERROR: Texture Manager failed to initialise!
Adding Global App Events
Registering System Sounds
Initialising Scripting Subsystem

Listening for commands...

Any ideas what can be wrong and how to solve this?

Thanks

I'd like to know this too! I've exactly the same problem. Anyone?

I received an answer from Blackmagic Support EMEA

Fusion is not designed to run on a virtual system and we don't support this type of configuration.
I'm not saying that it won’t work but as we don't test this I am unable to troubleshoot the problem.

I do hope virtual systems will be supported for at least the render nodes. Especially for the coming Linux version. More and more Virtual Systems are being used in large server farms. Please send a feature request to blakmagic if you also would like to see support for virtual Systems.

This brings me to an other problem. The licensing check being used for the render nodes. The version 7 render nodes could be installed without the need for a license check. They are running without a problem on our server farm. Now Version 8 needs a server with the USB dongle connected on the same network. Now this is very unfortunate for us:

1. The server managers do not like to see extra hardware connected to their servers. With virtual servers, ports are shared with other virtual servers. Without extra hardware the servers can quickly be moved to other hardware when there is a problem. Also, the dongle is very small and can easily be mistaken for an usb stick or get lost.
2. The server farm network is not connected to the company’s network. There are some strictly controlled firewalls but the security department will never allow us to open ports for this license check. This means I will have to buy an extra hardware license only for the render nodes.

I understand the need for a license check but using a hardware dongle is not the best solution for our server farm. A webservice would be a better solution for us.

For some of you it might sound strange why our render farm is on a separate network. We generate small personalized videos for marketing campaigns. With the render farm we produce thousands of personalized videos from compositions that are stored on a filesystem. We do not produce big movies so the user of Fusion Studio does not need this render farm.

What hypervisor? What vGPU do you have for the VM's?

Arthur van Rooijen wrote:The server farm network is not connected to the company’s network. There are some strictly controlled firewalls but the security department will never allow us to open ports for this license check. This means I will have to buy an extra hardware license only for the render nodes.

That's a business decision on your end, not BDP's. The solution you need is available and works fine.

I got the same problem on just native windows rendernodes. It just closes itself after a few seconds.
What I think is going wrong on my side is that my license sever runs on the wrong network. I don't know how to change this. I'm not familiar with environment variables. I read something in the manual about adding environment variables...Well.. I know where I can find them but I've no clue what I need to add to make the license sever use the right network?

Chad Capeland wrote:What hypervisor? What vGPU do you have for the VM's?

I need to check that with the server guys.

Chad Capeland wrote:That's a business decision on your end, not BDP's.

Yes, it is, but I'm sure that a lot of company's do separate their public network that is connected to the internet from the corporate network. It would be very unwise not to do so.

Chad Capeland wrote:The solution you need is available and works fine.

You say the solution I need is available and works. Do you mean having two dongles? One for the person using Fusion Studio and one for the render nodes. Or do you have an other solution?

Arthur van Rooijen wrote:You say the solution I need is available and works. Do you mean having two dongles? One for the person using Fusion Studio and one for the render nodes. Or do you have an other solution?

Your security department needs to do their job correctly. They're either blowing smoke or they don't know what they're doing.

Chad Capeland wrote:Your security department needs to do their job correctly. They're either blowing smoke or they don't know what they're doing.

I think they know what they are doing but they don't want to open an extra hole in several firewalls.

Arthur van Rooijen wrote:

Chad Capeland wrote:Your security department needs to do their job correctly. They're either blowing smoke or they don't know what they're doing.

I think they know what they are doing but they don't want to open an extra hole in several firewalls.

Push back. Their laziness is trumped by your business needs. If you need to, embarrass them, the tail shouldn't wag the dog.

We're seeing this on our dedicated (non-virtual) nodes with a simple loader and saver comp. to test with

Code: Select all

2016-07-14 17:06:16:  0: STDOUT: Initialising Class Registry
2016-07-14 17:06:18:  0: STDOUT: Initialising all geometry caches
2016-07-14 17:06:18:  0: STDOUT: Initialising GLTools
2016-07-14 17:06:18:  0: STDOUT: ERROR: GLTools failed to initialise!
2016-07-14 17:06:18:  0: STDOUT: Initialising Texture Manager
2016-07-14 17:06:19:  0: INFO: Process exit code: -1073741819
2016-07-14 17:06:19:  0: Render time for frame(s): 3.520 s

Anyone found a solution or answer to this yet? Is there binaries we need to make available to the node with the standard vga adapter inside it? (no GPU as you'd expect).

Cheers

Little Update:

I actually pulled one of the Blades from the rack, opened it up and installed a spare graphics card (GT640 w/ 2048 MB RAM).
Guess what- now Fu8.1.1 rendernode/and Fu8.1.1 Studio work, didn't even have to reinstall.
As this is not at all a viable solution i plea to blackmagic/the devteam that this issue must be adressed immediately.
Like highest prio with a hotfix.
As of now i think Fu8.1.1 is not usable for production, except when all your blades have gpus or you render everything locally.

Thanks Moritz,

We've been testing with all generations of nodes in our dedicated farm here and even on the latest dedicated nodes, with a simple Dos shell with the right envs set for licensing, can't even load FusionConsoleNode.exe to start a render.

A massive +1 for a hotfix that solves this lack of ability to render on dedicated nodes from me here.

Cheers

Jamie