Current state of Linux/Cloud deployment

[peedrr]

I have spent an embarrassingly large amount of time trying to get CasparCG running on Linux in AWS! My ultimate goal is to have CasparCG running in a cloud environment, ideally on Linux/Docker and without Nvidia GPU, hopefully allowing it to be run on any cloud provider, however I have attempted to build/run on both Linux and Windows within AWS, both with and without GPU instances, and nothing has worked.

Before getting into specific issues and errors, I thought it would be worth going back to the community and asking which of the following 4 scenarios are known to work:

Cloud VM (Ubuntu), native install

• with GPU
• CPU only

Cloud VM (Ubuntu), running within Docker

• with GPU
• CPU only

Cloud VM (Windows Server)

• with GPU
• CPU only

As well as experiences with the above, if anyone can provide instructions or system requirements for getting up and running within any kind of AWS/other cloud VM then I would be hugely grateful and more than happy to submit a guide to the Wiki.

Many thanks

[Julusian]

Current versions of Caspar require a GPU for compositing and CEF to work. If CEF is disabled in the build it might be possible to run without a GPU, with the help of some OpenGL CPU/Software renderer, but I havent tried doing this so can't confirm it.

The Cloud VM (native) with GPU should work, as it will appear to caspar as a normal linux machine, and that definitely does work.

I have successfully managed to run caspar in docker (https://github.com/CasparCG/server/blob/master/tools/linux/start_docker.sh), so I know that part is possible. One thing to note is that Caspar relies on being able to connect to a X11 server to be able to get the OpenGL context. So make sure to have that running somewhere, most likely on the host.

I can't comment on the stability or reliability of either of these, as I have only run it in docker as a bit of an experiment or on linux while doing development.

[peedrr]

@Julusian - thank you so much for your input and apologies for the slow response. Professional life took over!

One thing to note is that Caspar relies on being able to connect to a X11 server to be able to get the OpenGL context.

What are the current best practices for running CasparCG on a headless server in Linux? Back in the old forum I remember someone recommended using Xvfb. Is this the best practice? And if so, would Ubuntu Server be the best candidate (with their official Canonical releases available in all the major cloud providers) or is there a reason to run on Ubuntu Desktop, even if the ultimate goal is headless and/or cloud-based operation? If not Xvfb, which X11 server is recommended?

As before, I'm looking forward to updating the Wiki with best practices, just as soon as I find out what they are!

P.S. I am also constantly searching for info on this wherever I can find it - casparcgforum.org, cloned CGC github projects, etc. - but it's really hard to find current information now that the official CGC forum is dead (RIP) and the knowledge base has become so fractured. If you know of any thread or blog, etc. which has relevant info then I would be really happy to receive a link in here and I will be sure to link back to those resources in the Wiki. (Assuming that the official Wiki should still be the main source of info?) Thanks again

[chrisspiegl]

Hello @peedrr , I am wondering if you had any success on your quest of running CasperCG on a headless virtual server machine. I am thinking about the same and would love to see if there is an update.

I get that the current official requirement is still a GPU, however, I would be glad to cut some performance and quality (namely go down on the output resolution) if it would mean to be able to run on a sub 30$/m server.

Any info appreciated 🌸.
Chris

[nareshuppada]

@chrisspiegl @Julusian @mint-dewit @peedrr recently I was able to bring up the current version of casparcg on azure cloud with docker. It seems to be using cpu processing. Any luck on selection of cloud GPU for casparcg with docker.