Render Node and RAM

[glorybound]

I have issues with my render node, which lives on my 64GB RAM laptop. My Main PC has 128GB, and my average scene uses around 60GB of RAM. Would this cause my render to refuse connection? I can close this scene and open a smaller one, and it renders fine using the node. Even 60GB seems like a lot, and it never consumed that much RAM in previous V-Ray versions. What can I do to preserve RAM if this is the case? Everything that can be proxied is.

[vladimir_krastev]

Hi, thanks for posting.

Would this cause my render to refuse connection?

I assume you are using distributed rendering. Are you able to monitor your render node's hardware utilisation? Do you have any warnings or errors in your log?
If you are running out of ram it is likely that you render node will be slowed down a lot or crash. If it does not connect from the start it might be another issue.

What can I do to preserve RAM if this is the case?

Check your displacement, check your geometry and check your proxies. In rare cases a proxy can get corrupted especially if you merge very old scenes that were built with versions of V-Ray.
If you have bad topology of your geometry with displacement over it then you may have issues.

Try to go trough your scene locally first, try to disable layers, hide object's and so on to find what consumes the most memory.

If nothing helps, send us the scene. In this case it will be best if you use the support contact form. Please include a link to this thread so we can track you issue faster.

[glorybound]

Okay, I didn't think it would refuse if it didn't have the RAM; it'll just render slower. It renders without errors on the main PC. I'll see what I can find.

[Joelaff]

First, watch the RAM usage and see if that actually is the issue. It may be unrelated.

Also, note that the RAM usage under Max with the UI is typically a bit higher (sometimes a lot) than when rendering on a node. However, not sure how this applies to DR.

Second check your network settings. Disable any power saving stuff in the Adapter settings Configure section. We have found this to be especially important if using 2.5Gig ethernet, like the Realtek controller on so many new motherboards. Realtek is generally not the greatest, and their power saving and energy efficient ethernet settings when on (the default with the versions we had) would causes all kinds of issues with 10Gig switches.

But you really need to do some troubleshooting and see what might be the root cause. You are waiting long enough for the node to start up and get running, right?

[glorybound]

I have it on Performance, so no power issues. When the node doesn't connect, I just let the main PC do its thing, so yes, I wait long enough.

[Joelaff]

Performance won't fix the Realtek 2.5Gbit issue. The setting smust be set in the Adapter Settings->Configure. Of course if you don't have that NIC then it is irrelevant.

I meant do you wait for everything to startup when you setup DR on the node from the main machine? How are you starting the node? Manually? Deadline?

Check the log tab in the render window and see if there is any helpful info there about why the remote node is working. I do find DR to be rather flaky.

[glorybound]

The log quickly says that the render note actively refused connection. That is the only warning that the log says. My laptop is directly connected to my PC via Cat6. It works sometimes, but most of the time, it doesn't work. I am using DR in V-Ray,

[Joelaff]

Actively refused is a specific thing. It typically means the port is closed, or the firewall is set to refuse rather than silently drop. Most firewalls silently drop by default. That sounds like nothing is listening on the DR port.

Maybe this will help:

https://support.chaos.com/hc/en-us/a...shooting-Guide

[glorybound]

I can ping both ways, and the firewalls are off. I'll look at the guide; thank you!

BTW.. I have it set up as a service.

[glorybound]

For example, I am working on a scene where the DR was refused the last three renders, but on the fourth attempt, the render node worked. All I was doing was tweaking the camera view; nothing else in the scene changed. Nothing on the two PCs changed either. There doesn't seem to be a rhyme or reason for the slave PC to randomly not connect.

[Joelaff]

It could be that the remote node is recovering from the last render. On complex scenes this can take a while (like even a minute or more) before it is ready to render again. If you have restart between renders turned on in the distributed settings it takes even longer.

[glorybound]

It can be the first render of the day. If it doesn't work, I just let the main PC render, and that can be a couple of hours, which if the slave isn't going to jump in by then, it never will.

[Joelaff]

Wow... And the slave has been awake and waiting for a good 5-10 mins before you start that render? Maybe try running it manually instead of as a service. I dunno. Good luck.

[glorybound]

I have tried stopping and starting the service, but that doesn't matter. I can try turning off the service and try to manually start it. We'll see.