Start Folding on GPU without Website

[ersguteralbaner]

Hello together,
Iam new to this and I want to run a Docker Container with the FAH Client. The issue I currently have is that when I start the containers on my Setups the Folding doesnt start automatically on the GPU. I have to go to the Website and manually select the GPU and disable the CPU cores and also click the start buttom. I dont want to do that extra step. Is there a way to avoid that?

I saw these options for the config.xml file but these doesnt seem to work on the latest version:

<power v='full'/> → Ensures full usage.
<gpu v='true'/> → Enables GPU folding.
<fold-anon v='true'/> → Automatically starts without user intervention.

I appreciate any help.

Kind regards

[calxalot]

You need to have a persistent work directory for the client.
Once configured, it is possible for it to auto fold.

[calxalot]

See also https://github.com/linuxserver/docker-foldingathome

[ersguteralbaner]

You need to have a persistent work directory for the client.
Once configured, it is possible for it to auto fold.

Iam using a persistent work directory and have mounted it into the container. ( -v /path/to/foldingathome/data:/config \) Do I need to also define it in the config.xml?
Iam using the same Docker Image you sent. It still doesnt seem to auto fold and I also have to manually enable the GPU on the website.

[calxalot]

Everything you need for the linuxserverio image is in the examples in their readme.

For some command line control, you might like lufah.

https://github.com/kbernhagen/lufah

With lufah, you can create a group with zero cpus and all GPUs enabled and start folding.

Assuming your GPUs are supported by the container.

Note it is up to you to not oversubscribe cpu threads.

[ersguteralbaner]

Ive discoverd the client.db where the data gets stored about activating the gpus in the Web Userinterface. It looks like this.
{
"on_idle": false,
"on_battery": true,
"keep_awake": true,
"paused": true,
"finish": false,
"beta": false,
"cuda": true,
"key": 0,
"cpus": 0,
"gpus": {
"gpu:01:00:00": {"enabled": false},
"gpu:05:00:00": {"enabled": true},
"gpu:19:00:00": {"enabled": false}
}
}

The last part is interesting for me as there is where the gpu are being enabled. So my question is: Is it possible to enable all gpus without knowing the pcie Ids. Iam asking because I want to deploy alot of docker containers remotely on several setups and I cant figure those out easy and also dont want to use the WebInterface to enable those for each setup i will be deploying.
I want to have them enabled by default.

Hope this makes it clear.

[muziqaz]

GPUs are enabled, when compatible platform is found on the system by fahclient.
I would assume within a docker you have to have CUDA, OpenCL platforms set up for fahclient to see them and mark GPUs as enabled

[calxalot]

If the containers are identical, you can probably safely configure a client, quit it, and share the client.db.

I recommend you do not edit client.db directly.

[calxalot]

Do not share a client.db with any work units.

[calxalot]

Lufah can do what you want.

[ersguteralbaner]

Lufah can do what you want.

Even though the setups will not be on a local network?

[calxalot]

No, lufah makes a direct websocket connection. But don’t you have localhost access in the container when setting it up?

[calxalot]

I'm afraid I don't understand what you're trying to do.
Why don't you want to use web control?

I'm a novice with docker.
Maybe someone else can help you.

[arisu]

Lufah can do what you want.

Even though the setups will not be on a local network?

You can use lufah over SSH forwarding or over an encrypted VPN. Or anything else that lets you tunnel an unencrypted connection through an encrypted and authenticated one.

[ersguteralbaner]

No, lufah makes a direct websocket connection. But don’t you have localhost access in the container when setting it up?

No I do not. I want to deploy Docker images on Setups that are not owned by me. There are Hosts on a decentralized network called Nosana, where they basically rent out their idle GPU power, where I want to deploy the FAH Docker container.
I think I cant even use the WebControl for that if its not in the local network, can I?