04 - Setup docker-compose.yml

[ogaland]

docker-compose.yml if the file that will define all the services to run in the stack.

Quickstart
If python is installed on your host you can use the script generate_env.py to generate a minimalist config

Code: Select all

python generate_env.py 

If you can't run it, you can check the env/examples directory, 2 examples are provided, one for a windows config with a 4070Ti and one for a linux config with six 3060Ti. You will need to create a docker-compose.yml, from scratch or by copying one of those 2 files and tune it.

Basically 3 services category are defined inside :

1. The benchmarker (benchmarker) : This is a standalone service for testing your video cards and find the best settings (best hashrate, lowest energy use), if you use it you should stop all other services to not interfere with the benchmark results. When not needed it is better to stop it.

2. The watchdog (watchdog) : This service will monitor system health and pool progress, it provides a Web UI. Aim is to detect possible failure are restart containers if needed or at worst reboot the system if completly hung.

3. The workers (bitcrack-client-xx) : This service will run block scanning daemon for one associated video card, thus there will be as many worker as you have video card available.

Here are the points to check in this file :

- The benchmarker service should not be changed.

- The watchdog service env_file section should contain the name of the .env.watchdog you want to use for configuration

- The workers services should contains as many worker as you have video cards (just copy/paste if you need more), then tune the following section :

• service name : Use a standard bitcrack-client-xx where xx is a 2 digit number (gpu number starting from 00)

• env_file : Use the .env.gpu.xx file you defined for this video card

• image : use ogaland/bitcoin-puzzle-bitcrack:tag where available tag is picked from following repository Docker hub
  tags ares on the scheme ccxx_version, you should use the image on the latest version and with ccxx matching compute capability of you gpu, if you don't know it you can find it on nvidia site or by typing following command

  Code: Select all

  nvidia-smi --query-gpu=name,compute_cap --format=csv

  For instance a RTX 3060Ti is compute capability of 8.6 so the image to use is ogaland/bitcoin-puzzle-bitcrack:cc86_beta.0