CNCF Atlas

English

日本語

English

日本語

Appearance

Getting Started ​

Verified against release v3.5.3 and master commit 6b2e792. Commands assume Linux with Docker, or Go 1.18+ and gcc for a source build.

Prerequisites ​

  • Docker and docker-compose for the quick path, or Go 1.18+ with gcc (CGO is used by libsdk and blobstore) for a source build.
  • At least 10 GB of free disk for the DataNode; the script enforces MIN_DNDISK_AVAIL_SIZE_GB=10 (docker/run_docker.sh:6).

Install ​

Clone the repository and build from source:

bash
git clone https://github.com/cubefs/cubefs.git
cd cubefs
make build

make build produces cfs-server, cfs-client, cfs-cli, and the other binaries through build/build.sh.

A first working setup ​

The fastest way to a running cluster is the bundled Docker compose stack, which starts master, metanode, datanode, objectnode, client, and monitor together.

  1. Start the full stack, pointing it at a disk path with 10+ GB free:

    bash
    ./docker/run_docker.sh -r -d /path/to/disk

  2. To run a single role manually instead, point the binary at a role config. The role is read from the role key (cmd/cmd.go:184), and the dispatch switch maps it to a server (cmd/cmd.go:206-239):

    bash
    ./cfs-server -c master.json

Valid roles are the constants in cmd/cmd.go:71-93: master, metanode, datanode, objectnode, authnode, console, lcnode, flashnode, flashgroupmanager.

Verify it works ​

The compose stack starts a monitor web UI (the -m flag in docker/run_docker.sh). Once the stack is up, use cfs-cli to query cluster state, then mount through the FUSE client or send an S3 request to ObjectNode to confirm reads and writes round-trip.

Where to go next ​

For production topology, high availability, security hardening, and scaling, use the Helm chart (cubefs/cubefs-helm) and CSI driver (cubefs/cubefs-csi) and follow the official documentation rather than the single-host Docker stack (S1, S2).