Tweet
So, this is an issue I think I have brought up before, and so have other folks. I'm hoping I have something new to add, which may be helpful?? I'm using Vray version 3.60.03.
I am rendering a very simple scene. A bottle of pills. I can get a decent rendered result in 17 seconds when using DR, and maybe 30-40 seconds without DR. I have set up a script to render the Front, Left, and Right sides of the bottle, it then switches bitmaps for the label and repeats until it has rendered all the different labels. The script part is really not relevant though, as the issue occurs when rendering animated frames too. It's really anything that sends a command to DR to begin a new render immediately after completing a render. I have my local node rendering and also 3 slave machines.
What happens is that the first rendering finishes, and immediately sends the next image to DR. However, the slave machines don't often respond. The local machine will render this frame by itself, which is comparatively slow of course. Once this rendering is complete, it sends the next job to DR, and now that the slaves have had a short rest, they will all connect and contribute to this image. On the next job, they don't connect. Rinse and Repeat. Render, Fail, Render, Fail, Render, Fail. I have seen this on regular animation jobs as well, not just scripted render commands. Sometimes it will vary as to which slaves contribute on each frame. But they will almost always skip a frame and "rest", or at least have a significant delay until DR makes a 2nd attempt to reconnect (if it does). So at best, a slave will jump into a frame further along. I was seeing this happen with the local machine doing 80% of the frame, and then the slaves jump in for the final 20%. Pretty much a total waste of resources.
In the maxscript, I inserted some "sleep" commands in between each render command. Pausing for 2 seconds, 5 seconds, 10 seconds, 15 seconds.... all failed to improve the situation. However 20 seconds allowed ALL 3 slaves to reset and consistently contribute to every render job. It should be mentioned that I toggled all the DR settings- restart servers, save servers, transfer assets, etc.... None of those options had any affect. Nor does it matter if I use Progressive versus Bucket, GI on or Off, or any of that stuff.
Any thoughts?? Thanks!
I am rendering a very simple scene. A bottle of pills. I can get a decent rendered result in 17 seconds when using DR, and maybe 30-40 seconds without DR. I have set up a script to render the Front, Left, and Right sides of the bottle, it then switches bitmaps for the label and repeats until it has rendered all the different labels. The script part is really not relevant though, as the issue occurs when rendering animated frames too. It's really anything that sends a command to DR to begin a new render immediately after completing a render. I have my local node rendering and also 3 slave machines.
What happens is that the first rendering finishes, and immediately sends the next image to DR. However, the slave machines don't often respond. The local machine will render this frame by itself, which is comparatively slow of course. Once this rendering is complete, it sends the next job to DR, and now that the slaves have had a short rest, they will all connect and contribute to this image. On the next job, they don't connect. Rinse and Repeat. Render, Fail, Render, Fail, Render, Fail. I have seen this on regular animation jobs as well, not just scripted render commands. Sometimes it will vary as to which slaves contribute on each frame. But they will almost always skip a frame and "rest", or at least have a significant delay until DR makes a 2nd attempt to reconnect (if it does). So at best, a slave will jump into a frame further along. I was seeing this happen with the local machine doing 80% of the frame, and then the slaves jump in for the final 20%. Pretty much a total waste of resources.
In the maxscript, I inserted some "sleep" commands in between each render command. Pausing for 2 seconds, 5 seconds, 10 seconds, 15 seconds.... all failed to improve the situation. However 20 seconds allowed ALL 3 slaves to reset and consistently contribute to every render job. It should be mentioned that I toggled all the DR settings- restart servers, save servers, transfer assets, etc.... None of those options had any affect. Nor does it matter if I use Progressive versus Bucket, GI on or Off, or any of that stuff.
Any thoughts?? Thanks!
Comment