By dmeyer
#309497
1. How do I designate an IP while launching NetworkRenderNode via command line? The command from 1.7 doesn't work, and there is nothing in the manual about it.

2. Why isn't the Network Monitor reporting machine names anymore unless its a server with a FQDN? 1.7 did this.
User avatar
By juan
#309506
Hi dmeyer,
dmeyer wrote:1. How do I designate an IP while launching NetworkRenderNode via command line? The command from 1.7 doesn't work, and there is nothing in the manual about it..
Do you mean how do you select which manager will the node bind to when it is launched through command line? In that case:

mxnetwork -rendernode:[ip]
or
mxnetwork -rendernode:[name]
will work.
dmeyer wrote: 2. Why isn't the Network Monitor reporting machine names anymore unless its a server with a FQDN? 1.7 did this.
Theoretically we are following the same node machines detection than in v1.7, using a very similar version of the network libraries. We will try to reproduce it here.

Thanks,

Juan
By dmeyer
#309514
juan wrote: mxnetwork -rendernode:[ip]
or
mxnetwork -rendernode:[name]
will work.
Juan,

Thanks for the reply. The above works as long as the machines are on the same VLAN. In 1.7 it worked across VLANs as long as the IP was specified. We have built our renderfarm in accordance with this. Is there a way to do it in 2.0?

Update: Monitor will not communicate with Managers across VLAN's as well...
By dmeyer
#309534
Another Problem:

When running Manager on OSX, Windows Rendernodes receive notification of a new job order. Then they just sit there. No rendering, no network activity, nothing. Monitor thinks that they are rendering however.

When you remove a job from Manager using the Monitor, and then reconnect to Manager, the job is still there. It only clears it out of the Monitor's local view, it does not actually clear the job from the Manager. The only way to remove a job from Manager is to restart it.

Monitor on OSX crashes uncontrollably when trying to submit jobs if there are any pending jobs in the queue that were previously cancelled. This i suppose would not be an issue if the above bug was not present.

In the job submission wizard, the texture path selection dialog does not append a final "/" You must manually add it or you'll get texture path errors and your renders will fail.

It seems overall that the network rendering is as fragile as ever.

Not to be entirely negative, improvements seem to be:
1. Much more efficient on network traffic.
2. Love the Batch submission idea
User avatar
By juan
#309541
Hi again dmeyer,

My feeling is that most of the issues you report are related to the same thing, probably something about running the manager on OSX; it could be a very small thing but can be affecting everything. Since you have a very specific setup we could send you an special version that will help you to find out quickly what's happening (one of the biggest adventages of the new network system is that the diagnosis is much easier).

What happens if the manager is running Windows or Linux? Do you get the same problems?

We will try to reproduce your issues accross the VLAN immediately, we'll keep you updated.

Thanks,

Juan
By dmeyer
#309543
I took the Macs out of the equation on this try:

10-frame animation sequence, 4 nodes.

Frames 1-4 go fine, each node runs its first frame.

On frame 5, the node that grabs it can't find a particular mxi file referenced in the scene. Odd, considering it just saw it a moment ago for its previous frame. Also, this error causes all the nodes to just stall out. The renders forever say "rendering" in the Monitor, but nothing is going on with the nodes.
By dmeyer
#309544
juan wrote:Hi again dmeyer,

My feeling is that most of the issues you report are related to the same thing, probably something about running the manager on OSX; it could be a very small thing but can be affecting everything. Since you have a very specific setup we could send you an special version that will help you to find out quickly what's happening (one of the biggest adventages of the new network system is that the diagnosis is much easier).

What happens if the manager is running Windows or Linux? Do you get the same problems?

We will try to reproduce your issues accross the VLAN immediately, we'll keep you updated.

Thanks,

Juan

I have taken OSX out of the equation for now. All machines i am testing with are Server 2003. The only thing this has solved is the Monitor crashes slightly less often on Windows vs OSX.

Still though, we rely on a dual platform farm. Hopefully this can be sorted. In the mean time we are going to have to revert to 1.7 for production reasons. I will keep a couple of machines aside for testing 2.0 :(
By rusteberg
#309551
dmeyer wrote:Another Problem:

When running Manager on OSX, Windows Rendernodes receive notification of a new job order. Then they just sit there. No rendering, no network activity, nothing. Monitor thinks that they are rendering however.

It seems overall that the network rendering is as fragile as ever.
yes, experiencing same thing here. running monitor and manager on mac with nodes running on xp 64. "Scene data sent successfully", then nothing happens..... I even tried with a 78kb .mxs.... no dice.

yes, the networking seems VERY fragile. something that seemed to stabilize towards the end of 1.7. the ONE thing i'm not too happy about is that with a render node license, you can't run a .mxs through maxwell.exe... has to go through node!? whenever i was having network hiccups with 1.7, i would commonly run the .mxs on each machine separately and then merge back together. (more work, but lower risk of that nasty failure message...)

p.s. any finds on the manager/node conflict? thanks again!
By cwilliams
#326600
Seeing the same issue here.
Windows Rendernodes receive notification of a new job order. Then they just sit there. No rendering, no network activity, nothing. Monitor thinks that they are rendering however.

When you remove a job from Manager using the Monitor, and then reconnect to Manager, the job is still there. It only clears it out of the Monitor's local view, it does not actually clear the job from the Manager. The only way to remove a job from Manager is to restart it.


Maxwell Manager and Monitor on Windows server (64bit)
Maxwell Servers running on Windows 7 (64bit)


Craig

Hi Sam- sorry, I am on a PC.... other than install[…]

Hi, I'm having number of problems using MXS Refer[…]

Same here. I'm on iMac Pro - and it's pretty fine […]

Maxwell for Maya on OSX

+1 Waiting for this one also. Wanna try the textur[…]