Fiber release tags publish official container images to Docker Hub at nervos/fiber and mirror the same tags to GHCR at ghcr.io/nervosnetwork/fiber. Release tags are published as image tags with the same version, and stable releases also update the latest tag.
docker build -f docker/Dockerfile -t fiber:local .mkdir -p ./fiber-node/ckb
# Put your CKB private key in ./fiber-node/ckb/key before starting the node.
docker run --rm -it \
--name fiber-node \
-e FIBER_SECRET_KEY_PASSWORD='YOUR_PASSWORD' \
-e RUST_LOG='info' \
-v "$(pwd)/fiber-node:/fiber" \
-p 8228:8228 \
nervos/fiber:<release-tag>If you prefer GHCR, replace the image reference with ghcr.io/nervosnetwork/fiber:<release-tag>.
On first start, the image copies the bundled testnet config to /fiber/config.yml if the file does not already exist. To bootstrap from the bundled mainnet config instead, set FIBER_CONFIG_TEMPLATE=/usr/local/share/fiber/config/mainnet/config.yml.
The image also includes fnn-cli for administration tasks.
The RPC service listens on 127.0.0.1:8227 in the bundled configs, so publishing port 8227 from the container is not enough by itself. If you need remote RPC access, update rpc.listening_addr in your mounted config first.
If the node is already running in a container, the simplest way to use fnn-cli is to execute it inside the same container so it can talk to the default 127.0.0.1:8227 RPC endpoint:
docker exec -it fiber-node fnn-cli info
docker exec -it fiber-node fnn-cli channel list_channels
docker exec -it fiber-node fnn-cli payment list_paymentsFor interactive mode:
docker exec -it fiber-node fnn-cliIf you enabled RPC authentication, pass the token the same way as on bare metal:
docker exec -it fiber-node fnn-cli --auth-token 'YOUR_TOKEN' infoThe full CLI reference is available in ../crates/fiber-cli/README.md.
Stop the container and back up its complete mounted data directory. The new image can migrate the recognized old database layout in the same mount without closing V2 channels. Migration refuses unfinished local V1 channels or settlement before changing records or the database version. Use the old image to finish that work, then retry with the new image. Legacy third-party watchtower entries are not supported by this migration; keep their original protection running.
Completed V1 history is archived without deletion, and V2 signing sessions are preserved. Do not run an old image against the migrated store or restore stale channel data after new commitments have been signed. See the upgrade instructions.