So with Thinkbox Deadline 10 you can quite easily fire up a renderfarm on Amazon EC2.
It is pretty simpel when you use Max or Maya with a popular renderer, because they have those Machine Images already setup and you just request those licenses in their marketplace (usage based licensing).
They do not have a Linux/Blender machine image, but of course one can build his own. Deadline on the cloud is also free.
I am wondering, has somebody gathered some experience with that already?
Not with Deadline but with the Brenda scripts by James Yonan.
It is a bit of a hassle to set up but once you´ve got the hang of it it is great:
Here is a step by steb guide on how to set up an ami which is probably the most complicated part of the whole thing:
oh man, James’ talk about this is awesome.
checking this out, thanks
Hey Lumpengnom, I am wondering which region you are rendering on.
I am using Frankfurt-region eu-central-1, but the command
brenda-work -T frame-template -e 10 push
throws an exception:
boto.exception.SQSError: SQSError: 400 Bad Request
Googling down the rabbit hole reveals that this is because:
The Frankfurt AWS region (Ireland and CN too, apparently) only support the V4 signature algorithm.
and I need to add
[s3] use-sigv4 = True
to my ~/.boto file
I created that file, but that doesn’t solve it.
Do you have this issue? Have you found a way around it?
Thanks!
I am using the standard region which is US East (N. Virginia). I think i remember reading somewhere that it only works there but I am at the same time pretty sure that i´ve rendered sucessfully for testing purposes in US-West (Oregon).
The last time I checked - which was a couple of years ago - US East was cheaper than Frankfurt. That might have changed, though.
I have never had that boto problem.
ok, it works in us-east-1 (N. Virginia). Thanks for your help!
That .boto issue was because my WORK_QUEUE name contained dots. Replaced it with underscores and the issue is gone. Have not tried it in Frankfurt though.
Have you ever had the issue that an instance kept running in the background, and you couldnt see it?
In this thread user kesonmis says
…an instance that ran idle for 3 weeks and generated a 200 $ bill for me.
Now in the EC2 Dashboard, I can see the running Instances and Spot Requests.
I guess, if all Instances are terminated and all requests are cancelled, one should be fine.
If you check the dashboard and terminate instances there, you are fine… if you terminate all instances after they have spawned. My problem was that Brenda job finished and usually it means all instances are terminated. And I usually check that. But for some unknown reason one instance still sprung to life and ran wild and idle without any actual job to do.
I see. So I will do paranoia-checks on that dashboard!
I haven’t got any jobs running, therefore I can’t observe this yet, but is the cost display on the dashboard realtime? The Usage seems to be.
Well, this has never happened to me but I don´t trust the script and allways check in the dashboard if all instances are terminated and if ALL SPOT REQUESTS ARE SHUT DOWN.
If you terminate an instance without closing the spot request the instance might launch again.
On the other hand, if you close a spot request it also terminates the instance it launched.
It has never happened to me neither… until it did. And I always check in dashboard whether instances have been terminated and spot requests not pending… until I didn’t check it once. And it just happened to be that time when an instance was resurrected 
Heh, yeah, there is a risk. Did you try contacting AWS? A few years back when the brendapro forums were still open one of the users there - i think it was Todd himself - accidentially left a bunch of instances running and was supposted to pay some horrendous bill. IIRC he was able to explain to them that it was a mistake and he got the money back.
Yeah, I contacted AWS support and fortunately got that rogue instance bill discarded.
glad you didn’t have to pay that!
Did you have to proof that it was an accident?
I really like the aws dashboard. You can setup a budget and all sorts of alarms when you get close to having it used up. The cost explorer seems also nice. All sorts of statistics
You have to be bit careful with the billing info, though. Usually it takes a while, like a day or two, until all costs are displayed there.
I see, thanks. I’ll be careful in the first month 
Man, I love how I can scale the farm up and down in the dashboard, really cool.
Scaling in the dashboard? What do you mean?
well, scaling as in requesting new instances with
brenda-run -P -i c4.xlarge -N 5 -p 0.07 spot
and scaling down by terminating instances in the dashboard.
Not the most convenient way, but you have full control.
No, I just wrote that I had an instance running by accident and it didn’t do any work, just ran idle. Thay can look it up in logs, I gave instance id.
1 Like
Ah, I see. I thought there might be way to scale up via the dashboard as well.