Questions: License server hosted on EC2 / Best practices for setting up VRay render nodes on AWS EC2.

[uforis]

Hi,

I just want to get some feedback (to save us some time) before going ahead with what we are thinking of doing.

We previously had a bunch of machines running distributed as a local farm (one location). But we are in the process of splitting our farm across several locations in the next few months.

(1) Would it be viable / proper for us to create an AWS EC2 instance to host our online license server?
That way, each location can point to a static IP/port and greatly simplify our network structure for licenses.

(2) Is there any recommended documentation for setting up VRay Render Nodes onto AWS EC2 instances?
Along with having our farm in several locations, we want to look into spinning up EC2 instances if we need extra capacity for short periods of time.

We've been searching and could fiddle with this setup / kill time - but I'm hoping perhaps either Chaos Group /or this community has some insights to point us in a/some direction(s) before we begin. I'm eager to share what we find.

Thanks in advance.

[ivan.slavchev]

1 - You can set up a license server on a VM. Make sure the MAC address doesn't change from time to time since it can mess with its certificates. You can also use a single file on a shared location for all VMs if you need to change license settings centrally - https://docs.chaosgroup.com/display/...over+a+network

2 - It's best if the VMs are equally built in terms of CRU, RAM etc. Static IPs, or DHCP that maps IPs to MAC addresses, or a working DNS resolution is recommended so you'd know which machine renders what and will be easier to assign tasks. In terms of licensing - if you don't have enough licenses for the extra instances when you need extra capacity - you can get 1-month Render node licenses when you need them. Another option is to offload some rendering to Chaos Cloud when you need a project rendered fast - https://cloud.chaos.com

[Neilg]

Once you've got an EC2 server up and running and you've remoted in / fiddled around with how the whole thing works send thinkbox an email about setting up the render instances. It's not hugely complex, but also I did not do it myself. Budget your IT/most networking savvy person to spend a full day on it, maybe 2 to solve inevitable license issues, that should get you mostly up and running.


You'll have that main server which you switch on from your ews web portal, then you'll remote into it & launch max to submit jobs. spinning up render node instances is all handled within deadline from that main node. You pick the disk image and type of machine you want, then deadline handles the rest. Creating the disk image is a little time consuming.
Spend a little time reading and poking around here - https://docs.thinkboxsoftware.com/pr...ustom-ami.html

I have found thinkbox support extremely attentive and on the ball when it comes to all things with our EC2 account. They are currently helping us create a linux image that we run alongside the windows ones so we can render via standalone to cheaper nodes.

Edit: worth pointing out we set all this up pre-amazon/thinkbox partnership, so we have been manually creating instances. I just saw & remembered they created the 'AWS portal' which was supposed to make all this easier. i think it is supposed to allow you to submit the jobs from your local machine, we manually ftp all the assets to the cloud and relink once they're up there.

[uforis]

@ivan.slavchev Thanks for the quick response. We will be setting up our new licence server this upcoming week. We were actually in the VRay Cloud beta and absolutely loved using it. Our artists really wanted us to continue using it - but the pricing didn't make sense for our daily render reviews. And why we are setting ourselves up with something internal (using our severely under-utilized 10 node license) and only using VRay Cloud if we are in a pinch.

@Neilg We've been looking at render queue software and have in fact decided on Deadline10 - good to know that they have great support. Thanks for the link. We will contact them once we've got the foundation in place. From my understanding, Deadline10 will allow us to distribute render a single frame across all nodes /or the option to spread frames across individual nodes. I'm hoping the latter means we are less dependent on uniform hardware specs. Can't wait to get this all running.