[kami]

hi all

I constantly got network render problems. Sometimes it works, sometimes it don't. It helps if I restart every render node before starting a new render, but sometimes even this don't help.

This is the image I get:
http://www.deiters.ch/maxwell/network.jpg

all machines show that they are rendering bot none acutally is (as the processor activity shows).

what could be the reason for this, or how can I fix this?
It'd be so easier to hold my deadlines if I could count on network rendering. Sometimes I even start the rendering manually on several machines and do all the merging afterwards just to be sure ... But this costs a lot of time!

thanks,
kami

[brodie_geers]

If I understand what you're saying, I've had this same problem since 1.7 days (see http://maxwellrender.com/forum/viewtopi ... t=+network ).

Basically, you start a render and things go fine. But the rendering status stays as Rendering after you stop the job and the monitor won't let you use that node, because it seems to think it's still rendering. Only thing you can do is manually restart the node (restarting the nodes via Monitor doesn't work). Does that sound about right?

I've never been given a solution. The only workaround that I've found that works fairly well, is to never use the Stop button to stop a render via the Monitor. Instead, change the target SL to something less than the current SL and let it stop that way. That seems to work better for me.

-Brodie

[polynurb]

Kami i have also had these problems.. but luckily 2.5 update seems to work well here now.

before i could always only send one job, then the second would behave exactly as you describe.

see here> http://www.maxwellrender.com/forum/view ... 14&t=33994

there was no real conclusion to this, but 2.5 solved it somehow.

it would be interesting to see what the manager says, but it looks like the nodes never get the scene file and/or dependencies.
are you using a mapped network drive on the nodes?

for debug purposes, if it can be done in your case, i would connect all the nodes and computers involved to a single simple switch (no vlans etc.); only these pcs.and without internet.

then turn off all firewalls, try it again.

you could also try to change the default port range (on all pcs) for maxwell netrender. just to be sure there is nothing wrong there..

are your OS all xp64? the screehshot has aero but monitors says "windows 64" for nodes..
i use xp64 on all machines and lately i added a win7 machine as node and it also worked fine (manager running on xp64).
in fact it was very robust, running across 7 computers in coop, like 20jobs in a row..

it is really odd that this bug is still alive on some configurations... you should contact technical support to help you through this.

[frosty_ramen]

i continue to have this issue even in 2.5 not as often but as recent as last night.
again still no resolution just work arounds!

[kami]

thanks for your answers.
All machines are using Win7x64 and are connected to the same switch, which is connected to the router (as DHCP-Server). I could disconnect the internet and see if that helps, but I'm not sure if the firewall is to blame when it sometimes works and sometimes don't. As mentionned, restarting the nodes sometimes helps, so I generally do this before sending a new job, but on this case, all nodes were restarted before.
I am using a mapped netword driver (z:/) on all machines an let the final rendering be saved there.

It's not just that the rendering sometimes don't start, it often occurs that a node is suddenly disconnecting or beeing displayed as disconnected while it still is rendering or the merging process sometimes fails.

I know the network render has been improved so many times (as stated in the patch notes) so I supposed the issue was on my side ...
If not, I hope it'll be fixed very soon. It's way to risky to use the network render on short deadlines where you'd need it the most.
I'll try to investigate a bit more.

cheers,
kami

[Half Life]

My problem with network render is really the networks themselves -- they are cranky beasts that need constant tech support love to not turn on you.

I used to work tech support for a large university and I dealt with the networks on a somewhat part-time basis -- I've been wanting to teach a simple course on the Maxwell network render features but every time I start I find myself wanting to go into all the things that could be wrong with the networks that could stop the software from functioning properly... which of course is not really the point, but it is an unavoidable issue.

That is not to say the software is perfect, but I don't trust networks in general most of the time...

My preference is to get as much processing power in one machine as I possibly can, as one machine is easier to maintain than two -- and so on.

Not that helpful I suppose...

Best,
Jason.

[Bubbaloo]

Try using a UNC path instead of a mapped network drive path. I started using UNC path a while back and have had fewer problems. I remember reading that the team had made some improvements to the way Maxwell handles mapped drives, but I haven't done it that way in a while. It may not be the solution, but it's worth a try.

[polynurb]

just for the statistics.

i use mapped drives (X:\ on all machines)

and have not had any trouble (pre v2.0 and post 2.5)

kami,
you can also try with sending dependencies.. if your are in doubt about network shares working correct

[kami]

thanks again for the comments.
what do you mean by sending dependencies? I don't have that much problem with the network share generally. I never have any problems connecting to the server(win7x64-machine with mapped drive as raid0-volume) from any machines (except some mac laptops, but I don't use them for rendering purpose) and it runs quite fast (>100MB/sec).

I do use UNC-path sometimes and I'm not sure if it really helps, since it's not working 100% of the time eighter.

Since I'm not a network expert I have no idea what could cause these problems. So I tried to keep the setup as easy as possible. All machines are connected to one switch, are using the same operation system (win7x64) and belong to the same "Home Group". I did not install any firewall tools except for the windows firewall where I did not change any settings and always accepted all firewall requests from maxwell.

[polynurb]

what do you mean by sending dependencies?

means you don't need a mapped drive at all..

[fuso]

I'm also having massive problems with similar issues. Check out this post over in the network section, it might give you some
clues but I think we're all pretty much left with work arounds for now. Looking forward to the next beta release hopefully soon!
http://www.maxwellrender.com/forum/view ... 72&t=36022

[kami]

well it works if I manually restart all nodes and the manager. (not with the "reset all", but by quitting and restarting the node on every machine itself). not a very convenient workflow!

another question: if one machine restarts during the render process, it should continue to render after the restart, shouldn't it? this doesn't work at all ...

[fuso]

another question: if one machine restarts during the render process, it should continue to render after the restart, shouldn't it? this doesn't work at all ...

It should work in theory and it used to, but not in this release I'm afraid. The whole network should get a jolly good make-over.
And it's worse, as the other remaining nodes do randomly not finish the job, or at least do not merge the mxi. So all you can do
is to merge them manually and then resume when all your nodes are online again. If a node has to be restarted planned, then I
recommend to stop the job before and resume it in cooperative mode, which surprisingly works rather well.

Forget about those 'reset selected nodes' or 'reset the entire network' for now, it doesn't seem to do the job. But you don't
have to manually restart ALL nodes, just the ones which got stuck, restarted or caused trouble otherwise. In my case it's just
the odd one (out of 9). Well, let's hope it will all be fixed soon. Best of luck...

[numerobis]

And it's worse, as the other remaining nodes do randomly not finish the job, or at least do not merge the mxi.

Forget about those 'reset selected nodes' or 'reset the entire network' for now, it doesn't seem to do the job.

yes, i have these problems too. And the preview function is only working one or two times for me. After that i have to stop and re-add the nodes to get this working again... nearly unusable.

[fuso]

@ numerobis: I couldn't agree more with you and I'm 'glad' that I'm not the only one as I started blaming our IT and network
people already (and we all know what it's like when you lose their support). I'm so frustrated now and really hate only posting
nothing but this. Normally, I wouldn't care much about network rendering features but looking at the current render times for
still images when using a little bit more sophisticated materials, you have no choice but using a farm in cooperative mode!!

Luckily the prize tag isn't that dramatic, but still, is it reasonable to call the current version 2.5?? To be honest, I feel like I'm
back in beta times which I'm sure you remember, realising that we both joined back in 2005. Anyway, I really feel let down a bit
right now and hope this will be resolved.