2 nodes running separately

[BorisGoreta]

I have one LW instance running 1st node and a second LW instance running 2nd node.

First instance is rendering properly but the second instance does not utilize the node even though in network properties the node is visible and marked as available and ready.

[juanjgon]

So you have two Layout instances running in the same workstation, and you want to use two additional render nodes, one for each Layout instance, right?

I am not sure if this is possible currently with Octane, but I can ask to the developers about it.

-Juanjo

[BorisGoreta]

Yes this is what I want to do to share the workload. One instance uses 1 local GPU + 1 node ( because there must be at least 1 local GPU for Octane to be able to use the node GPUs also ). This instance can then do the render job while I work on the 2nd instance which utilizes the remaining local GPUs plus the 2nd node. This would be very convenient work flow. Otherwise I can't do both at the same time. The nodes have low spec hardware except the GPUs so I couldn't even load the scene locally on the nodes to use them to render while I work. 8GB only on them and the scene I am working now eats 28 GB in Layout ( but only 4.5 GB in GPU, go figure ).

[juanjgon]

The problem can be the total number of GPUs. Remember that you are limited to 12, so if the two render nodes plus the workstation have more than 12 GPUs in total, one render node is going to be disabled.

-Juanjo

[BorisGoreta]

I don't have a problem with this, because when I use only 1 LW instance with both nodes activated all GPUs are working.

[juanjgon]

Ok. If you don't have more than 12 GPUs in total, this issue can be a limitation of the network rendering architecture. I will ask to the Octane developers about it.

-Juanjo

[BorisGoreta]

There is a problem with node recognition. As you can see here both nodes are selected but only the first one works. To make it work I first need to select the second one, start rendering and then activate the first one for both nodes to start working.

[hdace]

This is a nice idea. I don't need it now, but I think I'll try it someday if Octane supports it.

[juanjgon]

Boris, one of the Octane developers thinks it should work. You have to make sure that you have configured different master ports for each instance. Can you try this?

-Juanjo

[abstrax]

If you want to run two masters on the same computer you must have the option "Auto-configure master port" enabled. The reason is that only one process can use a specific port. So if you disable the automatic port selection and use the same port (default is 21000) on two masters, only one master will get the port and the other won't and with that no slave <-> master connection can be established on the that master. If you enable the automatic port selection, that master will look for free ports and take the first free one it can find.

If you want to use both nodes, you will be using 10 GPUs, if your master uses more than 2 GPUs (makeing the total number > 12), one of the slaves will not be enabled. If you've got a log window in the plugin, you should see a corresponding error message.