PC renderfarm

[Thomo]

Hi Diego,

Please clarify what machine and OS (WIN 7 ??) you are sending the renders to Rama slaves from. I understand the slaves are i7 Win 7 (?). What is the master machine and specks please? Graphics card too.

 

I am sure you understand the point also being that it does not seem to be a slave side problem as it is fine if the same project is sent from a Mac OSX machine to the same PC slaves.

 

I use i7 and Win 7 Ultimate also, so that is not what causes the problem or prevents it. It happens from a Win 8 master machine also.

 

I am guessing, but were the projects you have been testing produced on your Mac and then tested from a PC Master to PC slaves? Have you tried to import / merge or create a scene solely on PC and render from Master PC to a PC renderfarm? 

And by the same point, were the projects you produced in Hamburg produced on a PC Master?

 

I have sent large projects to EIAS before where there are problems and they have not been well received. The problem being for me is that it happens more readily in a large scene, but not always. It is intermittent.

 

Put it this way, I have a scene that I can send right now that is doing it right now but it is large.

I'll post it later as a starting point, we might get lucky.

The up side of this "bug" is that you do not have to wait for a large scene to render to find the fault. That is the point, it does not render.

The down side it that it takes a minute or so to load.

[Diego]

Hi Thomo,

My main PC machine is a Intel i7 CPU under Windows 7 Home Premium 64 bits 16GB RAMS and graphic card an Nvidia Quadro 600. Same for the slaves except the graphic card.

All my projects are done in Mac. I tested with projects done in PC and the problem has not appeared.

Please try to simplify to the maximum the project with which you are having this error, try deleting items until the problem disappears, then returns a step behind and then upload it to DropBox or similar and well see if in my render farm is still producing the same problem, see if we get so isolate the problem.

For me the problem is not in the main machine or slaves or in any projects if not rather on a specific network or Renderama status. Let's see.

Thanks

Diego.

[Thomo]

Hi Diego,

Ok understood.

In light of your comments perhaps it would also be beneficial if you posted a project that is not having the problem for you but may for me.

Also if you have a project that was causing problems in Hamburg, but is not now for you, it may have problems with my network status. Let's try.

 

It does seem to be a clue if you have those same projects that were having problems in Hamburg but now are not on your "home" network. Let's try.

 

Let's to and fro between ourselves with some projects rather than bombard the EIAS crew with unnecessary information perhaps?

 

I am aware of the EIAS beta testing methodology.

 

I'll setup a dropbox link and we can both input to that perhaps.

 

Thanks,

Greg

[Thomo]

Hi Diego,

Just to clarify.

You have never come across this problem before in your office with a PC Master to PC Slaves? Ever.

[Diego]

Hi Greg, I've never had this problem in my studio because it is not in my workflow, I use Mac for scene creation and I find easier to use a Mac as master Renderama Machine.

The only time that happened to me was in Hamburg collaborating with the Planetarium of the city, developing an animation for full dome projection. They have all PC's and this problem about long time for "writing status" began to occur.

To my surprise when Tom asked me to reproduce the same error in my studio I prepared a PC as Renderama Master and the error did not appear in any of the ways.

So we are now waiting to see if the project that produce error in your render farm also produces in mine and begin to isolate the problem.

Thanks

Diego.

[Thomo]

Hi Diego,

Ok understood.

I also used to be Mac master based to PC renderfarm and never had the issue.

I am now all PC and am encountering the problem.

 

The main problem is re-producing the error. Once it happens, I re-start PC master and it goes away. Until next time when it appears again for no explainable reason.

I am working on it daily to isolate an instance but it is proving difficult for the above mentioned reason.

 

I am also going to setup a project to send up to Amazon as I was regularly getting the problem using that farm setup. This may also be a way of isolating the problem on a particular network that you and EIAS could use to replicate it. More on that later.

 

Please keep trying at your end as well. Also Meester Smith if you read this maybe you can contribute as well please.

 

Thanks,

[meester smeeth]

Hi Fellas

 

Yes our biggest problem here is as you say - re-producing the error.  I've been keeping an eye on it for a few months now and not seen it happen, but it usually happens here with a lot of repeated rendering and that hasn't been going on too much.

 

I don't even think we can rule out problems that could be caused by network settings or how they behave when idle for a period of time etc. The reason I say this is that I have a 2007 iMac and it will literally not work as a Master to Pc slaves, but works perfectly every time as a slave from a Pc master.  If they don't want to communicate from the start, something is in the way, from one direction only and nobody else has had this problem other than me!  I now have a 2012 iMac and will test it at some point.

 

Greg, is it mainly the Amazon set up that is doing it?

[Thomo]

Hi Mate,

Yeah it's a pain when it happens and then stopping it when it does happen.

Amazon works perfectly for me from Mac master, but PC master it seems to get the problem more often than on my own renderfarm. Having said that I am right now trying to make it happen on Amazon and it hasn't yet.

 

I'll keep trying and keep posting.

Cheers,

[Thomo]

OK.

I have a theory.

This is not a network problem, it is a graphic card problem.

I speak from an un-educated point on this.

 

My fear is this will not be able to be replicated unless it is on a particular WIN system.

My reasons and conclusions are as follows:

 

- If I have several graphic orientated applications open, Rama is more likely to falter.

 

QED: Rama (and EIAS camera window) struggle with large Rotoscope images. EIAS crashes regularly trying to refresh a camera window or a texture window (raytrace slot often).

WORK AROUND: Close other graphic orientated applications, set camera window to wireframe.

THEORY: Rama tries to refresh the input information from the slave graphically and cannot, so "goes to sleep".

FACT: Rama master window shows un-updated information window. However, close Rama master window (showing 20% of strip renders, for example, completed at that point) and then re-open, it will show actual strips or frames rendered at a higher percentage. (40 or 60% for example)

 

I have put together  problems with EIAS, CAMERA WINDOW and RAMA as I think it is all related.

 

If a graphic card is "overloaded" on a PC machine, rama cannot update itself and "gives up". I say this because Rama keeps rendering, displays less than it has actually completed in the Rama master information window, but after a re-set it displays what it has actually rendered. (as a percentage)

 

I have related this to the actual camera window in EIAS on a WIN system because it has serious "crashing" problems. I was going to treat this as separate issues but I feel it is related.

EIAS has trouble refreshing the camera window if other graphic using applications are also using the graphic card.

Therefore RAMA experiences the same problems.

 

As I said, I am not educated in these things but in my problem solving capacity this is the avenue to explore.

 

This will not be able to be replicated as a "bug" is my feeling.

 

Any thoughts anyone? Please don't ask me to send something and wait. 

 

By the same token, please tell me I am crazy, not right or a plain fool and I will test some more. Sooner rather than later please.

After I have eliminated this, advised by experts, as an invalid theory. I will test some more.

 

Any clarification of what I have tried to explain I will gladly provide.

 

Best regards,

[Thomo]

Forgot the conclusion.

 

I often have open on my machine simultaneously.

 

EIAS, RAMA MAster, FormZ, AutoCad, Photoshop. And perhaps; Revit and 3DS Max.

 

All are grapic "hungry" apps.

 

If I have ANY of these open and try to open EIAS, EIAS will crash. Close all and it will open.

 

Even if it is a EIAS file that was open before.

 

See what I am getting at.

 

So it is often a case that I have a FormZ model open, exporting to EIAS,  checking drawings with AutoCad, checking renders in Photoshop and sending updated renders to RAMA.

Then the RAMA problem occurs. 

In my simplified terms, RAMA keeps rendering, but when it gets no feedback (graphically cannot be displayed) it goes to sleep. Keep in mind I render in 10 strips for still images regularly. For the specific purpose that there is no other way to tell where RAMA is up to in the render process. Render as 1 image and wait and see.

 

I sincerely hope this helps.

[Diego]

Hi Greg,

To me it makes sense what you say, this explains that the problem does not seem to be related to a specific project, I currently do not have many graphical applications on my PC but I will do some testing to see if I can reproduce the problem so well. Developers are taking note of the tests are doing here. I will try to saturate the graphics processor and let you know if the same problem occurs.

Thanks

Diego