Best Practice Farm Setup

Render Manager General
#1

Hi Guys,

General question here.

I’ve been using Render Manager in combination with Scene Manager for a few months now. A few teething issues like everyone else, but enjoy the direction its trying to go. Looking forward to the new version to release soon which will hopefully deal with a lot of its shortcomings for specific management of jobs/nodes etc.

I’m just curious how the guys at Pulze intended this to work in an ideal scenario or any other studio for that matter. In the past, not to name specific farm management software used, I’ve set up the farm with a dedicated manager node. this stored the repository locally and all machines could access that server node. This server node never rendered, as it tended to cause problems with management with the CPU resources being bottlenecked by the rendering. I haven’t been running Pulze this way as I only had one machine and only recently added a second. Pulze has been notoriously flaky when rendering and the manager UI is up on screen.

So I’m preparing for scaling things up and intended to build a separate server node for just management again as I feel that would avoid a fare amount of instability in the management software (im also setting one up anyway to deal with other floating software licensing so its a two birds one stone deal). Is this the way you imagined it being used though or did you always intend for the main repository manager to be a render slave as well? What is the ideal way to set this up. I’m just aware that having a specific manager node will eat up a licence without being used for rendering. Not an issue in and of itself, its just not the way other render management systems have handled licencing with only render nodes themselves using a licence. It also means i don’t have to dedicate a super powerful workstation and can build something a little more mainstream and lighter on the pocket as well as install a server OS to get around the Window pro simultaneous connection limit. Not a problem now with the amount of machines, but will become in the future. I don’t want to have to start from scratch again to set this all up and would prefer to get it right while I’m in the thick of it now.

Look forward to hearing anyone’s thoughts on the matter.

:grinning:

#2

Hi Curtis,

Thank you for the detailed explanation of your current situation, and for your interest! In the following my statement will apply to the Render Manager 2.0.

So the long story short: Yes, we will recommend using a dedicated manager PC which never renders. This is the bulletproof solution for sure. If the manager PC also renders and it runs out of RAM then it is not recommended to use it for rendering and managing at the same time.

On the other hand using a dedicated manager PC won’t be necessary especially in smaller environments. So we developed a hybrid solution where each and every PC can be the manager automatically. This responsibility isn’t locked to a specific PC. In this case the manager status is flexible and the RM keeps running in extreme circumstances as well.

It is especially true if you have more than enough RAM for the render process. In this case there is nothing to worry about.

So my suggestion are the following:

  • If you are planning to have a small PC for holding licences then this is the ideal solution. Let RM run on it and the jobs will be managed flawlessly. Don’t forget that you get 2 additional free licences for RM.
  • If you have a powerful Node or Workstation with a lot of RAM then it is also good as a manager. So it can render and manage the jobs simultaneously.

If you have any questions, just let me know.

Best regards,

Attila

#3

As always thanks for the response @attila.cselovszki Its hugely appreciated.

I like bulletproof, commercially bulletproof is always the answer for me regardless how capable the machine is :grinning:

Its also futureproofing so it makes sense. I’m glad my planned path is still the way to do it and my knowledge is still relevant, therefore so am I :rofl: