I just started running v12 net render unlimited with 15 nodes and a dedicated server with two storage drives one for the project and models and the other just for the renders I call them drive X and Y. I am getting slaves banned from projects and the error states that it was unable to save to output path. The output path is to the server on to drive X. If I clear the banned machine it is then able to get on to project and render with out any issues. It feels a lot like file sharing issue when you open a project and drop a lot of nodes on at once.
Anyone else getting this? or does anyone know what might be causing this?
Thank you in advance on this matter.

Jack,
I have run into this as well, but just reset it and go.

I have no idea as the real cause, but think it may have something to do with some of the slaves trying to write at the same time.

The reason I say that is if you try a project that renders very fast (15 seconds per frame), you will get network timeout errors as well as the unable to save to path issue.

Just a wild guess........

David

As a workaround you could have the slaves save out to a local (to the slave) drive, and schedule a script to copy new frames to the network drive. That way, if access to the network drive fails, you'll still have a local copy and the slave can carry on with the next frame.

for each file in the local directory

copy file to network directory

if the copy succeded then

move file to local backup directory

end if

end for

Thank you for the reply and also to let me know I am not alone. This seems to be a pretty basic requirement for a render farm program so I am sure Hash can get this working correctly.
David,
I take it that you are seeing this issue in v13 as well?

Stuart
Thank you for the suggestion.

yes...

But like I said, I just reset it and go.

David

Have you filed a report on this issue?

No, it is not a big hurdle for me (right now at least).

David

I am wondering if this issue of banned slaves is not a net render bug but a server issues? I am using windows server 2003 the enterprises edition withe service pack 1 and I read on the other thread Rick saying server 2003 could only do 10 machines. I was under the impression that server 2003 was unlimited. I see the problem of banned slaves after going over 10 machines. I am looking into it today and will let you know what I find out.
Is it net render with the problem or is it our servers? That is the question.

Server2003 is accepts "unlimited" tcp/ip connection requests but only 10 at a time are open - the rest are queued.

Windows Workgroups, by comparison, only allow an absolute total of ten other PC's to permanently connect to a Windows resource. (This is not the same thing as a tcp/ip connection - the word 'connection" here is in the sense of having a mapped drive, or of loading a sharing a network printer into your printer list).

So the question is "Is it possible for a Renderslave to time out while its TCP/IP connection is waiting in the queue?".

I just do not know the answer to that.

Richard Harrowell.

Great question for Hash Richard.

"Is it possible for a Renderslave to time out while its TCP/IP connection is waiting in the queue?"