DR node sporadic behaviour

[tricky]

We are using 6 render nodes at the moment. 3 of them are 6/12 core i7 with 32GB Ram and 3 of them are 4/8 core with 16GB ram. Every so often, a couple of the 6/12 cores don't receive the DR job and just sit there doing nothing. Restarting spawner (running as a service on all machines) makes no difference.

I can't work out why this is happening - vray messages just says: warning: Render host xxx is not responding.

In 2013, why do I still have DR issues? Its very frustrating.

[vlado]

Do you use the option to restart the slaves after a DR render? Does this happen when you start a DR render immediately (or shortly) after another one?

Best regards,
Vlado

[tricky]

Do you use the option to restart the slaves after a DR render? Does this happen when you start a DR render immediately (or shortly) after another one?

Best regards,
Vlado

I do have that option enabled. It does seem to be when I start a render shortly after one has finished (or been cancelled) though I can't be 100% sure that this is the reason every time.

[vlado]

I do have that option enabled. It does seem to be when I start a render shortly after one has finished (or been cancelled) though I can't be 100% sure that this is the reason every time.

Ok then. It might be possible that the servers have not finished restarting if you start another DR very soon after the previous one.

Best regards,
Vlado

[tricky]

I will monitor it. Why can't spawner handle things like this better? It has always been a bit 'unsteady' when renders are cancelled or restarted.

[tashko.zashev]

The render slaves need to close the preview job before any new render to be started. The delay required between two renders depends on the scene complexity, slave hardware, network, e.t.c.
So it is important to let all of the slaves to finish/cancel the job before sending another one.

[tricky]

OK, but why can't some sort of logic be built in where if the particular server isn't quite ready, it will pick up when it is ready? It seems a bit 'old school' to try and second guess enough of a cooling off period.

[vlado]

OK, but why can't some sort of logic be built in where if the particular server isn't quite ready, it will pick up when it is ready? It seems a bit 'old school' to try and second guess enough of a cooling off period.

There is such logic, actually. Why it's not kicking in in your case, I'm not entirely sure. When a server starts up and is ready, it will broadcast a UDP message, which the render client can recognize and automatically add the node. This usually works, but if you have firewalls or other software that stops the UDP message or if it just gets lost, there isn't much to do.

Best regards,
Vlado

[victorwol]

just curiosity, why would one want to restart the server after each render?? doesn't that make renders always to take longer to start?? What is the benefit of that option?

Thanks

[vlado]

just curiosity, why would one want to restart the server after each render?? doesn't that make renders always to take longer to start?? What is the benefit of that option?

It does make the servers start a bit slower, but sometimes things remain in 3ds Max from previous scenes and things may get messed up.

Best regards,
Vlado

[tricky]

I'm just looking again at our 3 machines that seem to have this random behaviour with distributed rendering.

I notice that the active network is described as an unidentified public network. I'm pretty sure this should be set to 'Work' or a 'Domain Network' (we are on a domain), but I can't seem to find a place to change this. Any tips?

[svetlozar.draganov]

Probably from here:

[tricky]

Yep - found that, but it's greyed out (i.e. you can't change it)

[trixian]

I have had this problem before, specifically after setting up some VPN stuff. Seems windows get confused by this sometimes.

Another idea is to check if you possibly have any new network devices added into your device manager. Win update added several ip6 virtual adapters here, and they needed to be disabled, or at least that was the workaround I had success with.
Right-click the device manager entry in the computer management window and enable "view -> Show hidden devices". It should then show you a bunch of stuff you never knew you had.
Disable anything related to ip6 (unless you have a ip6 configured network, then I have no idea).

You might need to restart/relog to get it going.

[tricky]

I have noticed an item in device manager that has an explanation mark next to it. No idea what it is. There are no cards/boards installed on these computers apart from a basic nvidia gfx card. I did try a while ago to find out what it was but failed. The boards are Asus P9X79 with the latest BIOS installed (version 4302).

I have also, as you suggested, told it to show me hidden devices. There does seem to be something relating to IPv6 called WAN Miniport. Should I disable this?

These systems are very basic W7x64 installations. Nothing special at all. Just fast CPUs with lots of RAM.