srbminer-docker-nvidia

SRBMiner-MULTI for NVIDIA GPUs, built from the upstream release binaries.

ghcr.io/runonflux/srbminer-docker-nvidia:latest
ghcr.io/runonflux/srbminer-docker-nvidia:<srbminer version>   # e.g. 3.7.8

A workflow checks for a new SRBMiner-Multi release every 6 hours and publishes it under its version tag and latest. Each download is checked against the md5 published with the release.

Environment

Variable	Required	Description
ALGO	yes	Algorithm, e.g. pearlhash (Pearl / PRL). Full list: --list-algorithms
POOL	yes	Pool host:port
WALLET	yes	Wallet address (some pools expect address.worker here instead of WORKER)
WORKER	no	Worker / rig name, passed as --worker
PASSWORD	no	Pool password, passed as --password
CPU_MINING	no	true to also mine on the CPU. The default is GPU only (--disable-cpu)
EXTRA	no	Extra SRBMiner arguments, space separated, e.g. --tls true --api-enable

All parameters: https://github.com/doktor83/SRBMiner-Multi/blob/master/Parameters

Run

Needs the NVIDIA container runtime on the host.

docker run -d --gpus all --restart unless-stopped \
  -e ALGO=pearlhash \
  -e POOL=<pool host:port> \
  -e WALLET=<your PRL address> \
  -e WORKER=rig1 \
  ghcr.io/runonflux/srbminer-docker-nvidia:latest

Notes

• SRBMiner uses OpenCL for NVIDIA. The image ships the OpenCL loader and NVIDIA's ICD entry, but no driver libraries. Those come from the host through the NVIDIA runtime, so the host driver version is what counts. Some algorithms need a recent driver, for example noid needs 580+.
• SRBMiner is meant to run in a terminal (other SRBMiner images set tty: true or start it inside screen). Kubernetes and docker run -d don't provide one, so in that case the entrypoint runs it under script to give it a pseudo-terminal, and log lines end in \r. Setting tty: true (compose or Kubernetes) or docker run -t makes the entrypoint run it directly.
• SRBMiner charges a dev fee per algorithm (shown by --list-algorithms).