Auto add dr node to current render

[3LP]

Hi there,

I'm in the process of creating my little online render farm with AWS EC2.
I've got it all sorted out pretty well for animations rendering, but I would like to push it a step further for single image rendering.

I would like to be able to start a DR job on the master node and then be able to lunch instances on the fly and add them to the render.
My only concern is to have all the ips of the nodes before the actual render start.
Because the master node reads the vray_dr.cfg before the render starts and call all the nodes that are listed in there, I would need to have all the nodes ip before the render starts.
Because I don't know the ips before they starts, I can't add them in the vray_dr.cfg before hand.
Is there a way to interact with that file as the render is already started and hot plug them in the render as the render is already going?

Thanks

Stan

[3LP]

I guess this is too much out of borders to be considered?
Maybe this could be part of the new Node Manager that you guys where considering at some stage?

Stan

[vlado]

For the moment this is not possible; the V-Ray 3.0 SDK does allow adding of render nodes on the fly, but there's no UI to do that while 3ds Max is running on the client machine. Plus, if the client machine is running a backburner rendering, there is no UI at all... You are correct that maybe this is a job for a larger tool, but for the moment we don't have resources to focus on it.

Best regards,
Vlado

[3LP]

All good,

Thanks for your reply Vlado,

As a temporary solution, I was thinking of having the Master Node just reading every X seconds/minutes the vray_dr.cfg and try again to connect to all those computers, the same was as it already does. For the moment, if you have a computer listed but not running the spawner, it will jump in the render if you start the spawner. I deduct from this that the Master Node is just trying in loop to connect every x seconds to the computers that are listed in that file.
The only change that I thought would be that the Master Node was just reading that file again. This means that we could add nodes on the fly. No matter if we are running through BB or not, I understand that it would be easier for everyone with a UI, but I don't really need one...

Afterwards, creating that file (vray_dr.cfg) manually or automatically is easy, but that doesn't concern VRay, not in the first stage.

Stan

[vlado]

It is not impossible, certainly... will have to think about it a bit.

Best regards,
Vlado

[cheerioboy]

Can you use the computer names in the cfg instead of ip? I'm not sure how AWS EC2 works though..

[vlado]

You can use computer names, yes. Worked fine with EC2 last time I tried it...

Best regards,
Vlado

[3LP]

Yeah,
Well, even with the computer names, they are created automatically, so it wouldn't't really help, isn't it?
But it's easy to create a script at startup that send the current ip of a computer to a db. After, the Master Computer only needs to reed every x seconds that db and compile the vray_dr.cfg file and done...
I was thinking of doing it like that to make the whole process automated, but like I said earlier, someone else could could find it also usefull to do it by hand.
In the end, the whole point is just having the Master Node reading that file every time he times out the spawners in stead of only once when the render starts...

I would love you to have a look Vlado

Thanks

Stan

[3LP]

Hey Vlado,

Did you had any chances to have a look at this already?
I'm just on a project for the moment that could need some rendering power on demand to the cloud and EC2 would be the perfect match.
Although, the ip problem is still a issue to distribute render and I would love to have this feature added.

Thanks

Stan

[Sebastian Ek]

All good,

Thanks for your reply Vlado,

As a temporary solution, I was thinking of having the Master Node just reading every X seconds/minutes the vray_dr.cfg and try again to connect to all those computers, the same was as it already does. For the moment, if you have a computer listed but not running the spawner, it will jump in the render if you start the spawner. I deduct from this that the Master Node is just trying in loop to connect every x seconds to the computers that are listed in that file.
The only change that I thought would be that the Master Node was just reading that file again. This means that we could add nodes on the fly. No matter if we are running through BB or not, I understand that it would be easier for everyone with a UI, but I don't really need one...

Afterwards, creating that file (vray_dr.cfg) manually or automatically is easy, but that doesn't concern VRay, not in the first stage.

Stan

This is a very good idea!