Hi all, been using this site quite a while but this is probably my first post.

And it's a cry for help!

I recently moved up to Carrara 7 from 6 and wanted to start taking advantage of network rendernodes.  The setup was supposed to be easy enough though I had to disable Windows' built-in firewall to get the broadcast sent out / picked up by the one node I had ready to go.  Seemed even specifying the exact ports (5020, 5040, and 5060) didn't help.

I had the main hub and the node talking as the node loaded the scene, calculated particles, calculated lights and shadows but then nothing happened.  The CPU usage on the node went from 25% to 0%, and the mem usage went from over a GB to just 150k.  The main hub continued to render but it was clear only the four local cores were doing any work on the video frames.

I tried different ports, but no good.  I tried both wireless and wired connections, no difference.  The only other thing I could think of which I cannot change is that the hub is Vista and the node is XP.

When I eventually stop the render, a message appears saying it could not communicate with the node on port 5060.

So I'm not sure what to do next.  If any of you Carrara masters can assist I'd appreciate it.  And I'm happy to share more info if anything is useful.

Ta guys, I'll share the action-packed video with you on this thread if you help me fix this. ;)

Matt

Did you try putting the daz installation folder as an exception in your firewall settings?

Well I disabled it entirely on both machines to be sure.  Once it works I'll work it back up again with exceptions and opening ports. :)

I have just found out that it may be because I'm relying on connections through the router, which may not facilitate a broadcast domain like a switch would.  A friend is coming over tomorrow with a machine, extra cables and a switch box and we're gonna have play and see if this does the job...

Don't forget to look into your router settings. Port forwarding or triggering would be where you should start looking if you haven't covered that aspect already. It'd seem logical if those settings only affected things going to/from the internet portion, but with some router models they also affect the LAN too.

If you can access a network drive on one computer from another and vice versa, then you should also be able to get your nodes working. Good luck!

Ack, this confounded thing continues to elude me.  Ta for the responses so far.

I have my main machine hooked up to a gigabit switch now.  Also connected to that is an ESXi server running one machine to act as a rendernode.  I have two laptops here, one which is configured to be part of a domain at work (i7 CPU) and one which is purely mine for home use (i3 CPU), which I would expect to work better on my home network.

My main PC resumes rendering, with network rendering on and network logging on.

Here's how each of them reacts.

The i7 work laptop starts loading the scene, calculates lighting and particles, and then stops doing anything.  Mem usage is still over 800k sometimes, sometimes when I do this it drops to 150kish for Carrara.  CPU is 0.

The i3 home laptop does as the i7 does and this one seems to almost work.  Only when this laptop is running a rendernode do I see four grey 'N' blocks appear on the frame being rendered on my main Carrara installation.  However, they never render.  The i3 CPU usage for Carrara is at 0% while RAM usage is stuc at about 882k.

The ESXi machine also loads the scene, calculates everything, and then behaves exactly like the i7 laptop.

I have three machines here more than capable of working with my my PC's Carrara renders and it seems none of em want to talk to eachother.  I have set all the Render Room settings manually to match their IP addresses and all ports are reflected in eachother (defaults left as they are at 5020, 5040 and 5060).

What can I try to fix this?  It's so frustrating when it should be so easy. :(

EDIT: oh, this may be useful, the network.log contents:
Start Rendering:
 IP Address: 192.168.1.11
 Subnet Mask: 255.255.255.0
Broadcast to: 192.168.1.255
Registering Render Node: Ainge (192.168.1.12) 4 CPUs
Registering Render Node: bignode (192.168.1.7) 2 CPUs
Total Nodes: 0
Total CPUs: 6

I notice Total Nodes is 0, even after detecting the other two... is this expected?

Could this be a UAC problem on the VIsta side?

Can you turn off UAC and see if that fixes it?

Open up Control Panel, and type in “UAC” into the search box. You’ll see a link for “Turn User Account Control (UAC) on or off”:

On the next screen you should uncheck the box for “Use User Account Control (UAC)”, and then click on the OK button.

You’ll need to reboot your computer before the changes take effect.

A nice thought.  The OS's in use are
Main machine: Vista
i7 Laptop: XP
ESXi VM: XP
i3 laptop: Windows 7

Win7 has UAC still on, but the vista machine has had UAC off for some time now.  I could give the UAC disabling a go on the i3 machine but I'd have thought at least the ESXi machine would work?

I havent tried a mix of Windows 7 & XP with Carrara with Render Nodes.

Don't forget that all of your Carrara plugins, and I mean ALL, have to be installed on each render node. I'm not sure if that would stop the render from starting, but you would end up with dead blocks or lost pixels whenever a node takes over a spot where it needs it.

Also, you don't have any screen savers or power saving features turned on on any of your render nodes do you? If one takes a while or times out, it can pretty much hang the whole setup as it goes into power saving mode.

There are so many little stupid things that can bite you. But provided you get enough nodes running, the time it saves can be worthwhile.

Hi guys thanks for the ideas so far.

I have been monitoring the active nodes like a neurotic mother when I start rendering cos I just want to see it work :)

Strange thing, last time I left network rendering on (accidentally, I'd given up for a while) I came back after a while to see that the ESXi box had actually been rendering some of the image!  Thing is now I see it is unworkable because I've deployed a lot of camera-facing JPG texture particles and it seems different machines render them slightly differently from eachother.  I could actually visibly see the squares which had been rendered by the active node. :)

I tried something else - installing Carrara on these boxes and rendering a test frame of the scene.  Netiher box could manage to render a single image, encountering a memory allocation error (even with spooling way down to 10-20mb and the exe's laatido'd).  I thought the i7 laptop would have been more powerful but I'm starting to think the scene is just doing too much for these other boxes that aren't quite a supercomputer.

I'll try again once this scene is over with with something simple.  One of the preset canyons or something, and I'll see if the collapse of CPU processing is actually due to my ambitious animations.... :p

Ta chaps,
Matt

Good Idea Matt,

Let us know how it works out.

If I have to I'll setup a quick rig here with a mix of XP and WIndows 7 to test.

All my machines are quite old.  Slowest P4 HT and faster Core 2 Duo Laptop.

It's a known bug that if you render using GI, you will get different shaded results on different machines resulting in checkboard patterns in your render.  This was REALLY bad when C first came out with Render nodes.  It's a little better now.  I have no idea what it's like to try rendering on two identical machines (hardware and OS).  I've never been rich enough to have two identical machines at one time.

Hello, I thought it best to let you know that I setup a new less complex scenario and the network rendering is functioning very happily!  That's from a Vista 64 bit machine to two XP 32 bit OS machines, one with 2 cores and the i7 with 4.

I guess my first scene was just too much for them to cope with...

Thanks for your advice while I wrestled with this one.  When the event I'm doing this for is over I'll post a link ;)