[Micha]

Today I started to use the Maxwell network rendering, but I have several problems:

- some times if I try to get an image of the coop per "display", the image create process freeze and don't show an image (the single server "display" works) - it seems to be, if SL update and display are started at the same moment, it cause the freeze

- if a server crashed and the process is stoped, than the complete coop render process can't removed, because some communication processes are waiting - it helps to cancel the render manager

- some times I can't start a coop rendering, the start processes of the servers are freezed. Only close all servers and the manager helps.

The networking rendering could be a good feature to speed up Maxwell renderings, but here I got crashs very often. Some times the server process crash, sometimes the whole machine. I wonder me, because the last weeks I used an other render engine in DR mode and the slaves was working stable for many hours.
Is it a known Maxwell problem, that server processes and computer crashed often? Or should it work stable (all machines XP64)? I could try to reinstall Maxwell at all servers.

[tom]

Hi Micha,

It's right for several reasons the networking feature is not robust all the time. Freezes are possible with display calls as you say, we're trying to enhance this area. As it's a multiplatform system we're relying on some networking functions which we can't easily avoid problems with them. For any reason if a server in a cooperative session crashes, I'd suggest right clicking on it and canceling the job so at least it will be possible to merge the rest of the job without problems. Terminating the manager shouldn't be necessary unless the queue has completely gone non-responding and it's something quite rare. When something has went wrong with a server and it crashes, next time it's possible this issue could disturb the render manager and/or other render servers and prevent them doing their jobs without problems. So, when this happens I'd suggest restarting server sessions before submitting any job to the queue. For diagnosing crashes, please follow these steps below:

- Does it happen to same machine(s)?
- Does it happen with any scene?
- Does it still crash when you run the crashing server alone?
- Do you have enough RAM to render and what happens with a very small scene?
- Do you also have BSOD and could it be because the CPU is overheating?
- Are you using same version on all the nodes?

Also, provide more information about your system, hardware and the crash if possible...

Best regards,
tom

[Carl007]

Similar problems in OSX environment. It is sad, because I like Maxwell Render, and I would love to use it more in production, especially animation.

Could anyone please explain to me why Maxwell networking in OSX is so hard, I am not a sysadmin at my work, by the way, and not a programmer, but are there not some kind of OS system libraries that are used to pipe in network support in 3rd-part applications? Please forgive my ignorance.

[pwrdesign]

We've just upgraded to 14 servers for rendering and I think that the Network render works pretty well. We're running Linux64 on all nodes except the Manager (will switch that one to linux soon).

One thing I've found though is that its good to restart the MXCL server sometimes, Especially if the servers havnt rendered for a while.

I sometimes get "Securing server" which forces me to restart the servers and the manager.

Micha I had the problems you described before, when we was running win32 on the render-machines. All of them have 4GB of ram, but when we switched to Linux64 95% of those crashes stopped.

Maxwell is very RAM-consuming, and my experience is that if your number of emitters, ouput resolution starts to climb, together with ML, you also need to buy some more memory.

Atm We have 4 Gb of ram in all nodes, and it works great.
For ML adjustments and Sinulens I'm always doing this from one of the nodes with 8 GB of ram.

Regards Patrik

[Micha]

Today I found some time again ...

All four machines are using XP64 and the RAM is 8/4/4/3 GB. The scene is very simple - a glass simulation with 150MB RAM usage at the taskmanager. It dosn't crash at all machines, only slave1 (4GB) and slave3 (3GB). Sometimes Slave3 crash complete with shutdown and restart, at Slave1 the server exe crashs only.
All machines show me Maxwell 1.6, but I installed 1.6.1 at all.

It's a pity, that the "Displayed" image from all nodes can't be saved during a rendering is running. It could be nice to save a rendering and let it bake longer and if it should crash, than the saved image can be used. Or better, an automatic save of the rendering to the hdd every 10 min.

Interesting for me is, if I let start manager and server per "autostart" at slave1, than I see an error message like the attached. If I start the machine without server and manager and manualy start manager&server, than I don't see the message. Yesterday I have seen a start error message at slave3 too, today not, but the server crashed today too.

I don't use Maxwell often. Since a long time it's my first Maxwell project. In the last weeks often I used Vray network rendering (DR) and it works very stable. So, overheading should not be the problem, but ... now I set a much higher fan speed and no crash since one hour - a total SL20 is reached.

[tom]

All machines show me Maxwell 1.6, but I installed 1.6.1 at all.

It's OK, that's something forgotten in releasing 64 bit executable. The modification stamp should be 12 Dec 2007 Wed, 14:24:04

Interesting for me is, if I let start manager and server per "autostart" at slave1, than I see an error message like the attached. If I start the machine without server and manager and manualy start manager&server, than I don't see the message. Yesterday I have seen a start error message at slave3 too, today not, but the server crashed today too.

So you mean autostarting the manager and the server on Windows startup?

I don't use Maxwell often. Since a long time it's my first Maxwell project. In the last weeks often I used Vray network rendering (DR) and it works very stable. So, overheading should not be the problem, but ... now I set a much higher fan speed and no crash since one hour - a total SL20 is reached.

It only means Maxwell burns the CPU better and that's something good

[Micha]

Interesting for me is, if I let start manager and server per "autostart" at slave1, than I see an error message like the attached. If I start the machine without server and manager and manualy start manager&server, than I don't see the message. Yesterday I have seen a start error message at slave3 too, today not, but the server crashed today too.

So you mean autostarting the manager and the server on Windows startup?

Right, I copied the link symbols (I don't know the right english name of the WINDOWS symbols) of both to the autostart menu entry.

It only means Maxwell burns the CPU better and that's something good

Today, the house heating was not right working and wife was surprised at my good room temperature ...

[tom]

Right, I copied the link symbols (I don't know the right english name of the WINDOWS symbols) of both to the autostart menu entry.

This shouldn't be a problem as I do the same here without problems. Do you use any other resident software like an antivirus or firewall which scans startup applications on load? Have you tried doing this only for server but lunching the manager manually?

[Micha]

There is not antivir software on this machine, but today I copy the start entries for manager and server at the "autostart" for tests again and it works - no error message.