Run IPFS in a Docker container
by Kyle Drake on 2015-07-11
In recent years, Docker and a few other projects have redefined how we run server applications. In the future, we might be running containerized apps in our personal devices. At its core, this fast-paced improvement is a combination of good interfaces to standardize how to do things, and great tooling to make using containers easy.
The IPFS Project has many things planned for the world of containers. The most interesting is using IPFS to distribute containers hyper efficiently across data-centers and the internet. We will be discussing many of these things in upcoming posts, but first things first. This post is a quick guide for running an IPFS node directly within Docker.
The IPFS team has provided an IPFS Docker image, which is syncronized with the latest commits to go-ipfs. It only takes a few commands to try it out!
> mkdir /tmp/ipfs-docker-staging > mkdir /tmp/ipfs-docker-data > docker run -d --name ipfs-node \ -v /tmp/ipfs-docker-staging:/export -v /tmp/ipfs-docker-data:/data/ipfs \ -p 8080:8080 -p 4001:4001 -p 127.0.0.1:5001:5001 \ openthings/go-ipfs:latest faa8f714398c7a1a5a29adc2aed01857b41444ed53ec11863a3136ad37c8064c
8080 is the HTTP Gateway, which allows you to query ipfs data with your browser (see this example), port
4001 is what swarm port IPFS uses to communicate with other nodes, and port
5001 is used for the local API. We bind
5001 only on
127.0.0.1 because it should not be exposed to the outside world. The
faa8f7143... is the docker container id.
We’ve mounted a data and staging volume. The
data volume is used to store the IPFS local repo (config and database), and
staging is a directory you can use for staging files for command line usage (such as
ipfs add). If you’re only using the API, you can omit the staging directory volume. And of course, feel free to put those directories somewhere other than
Now what? Your node is running. You can issue commands directly to the containerized ipfs with
docker exec <container-id> <ipfs-cmd>. For example, you can try
ipfs swarm peers to see who you are connected to:
# let's set $cid = <container-id> for easy access > cid=faa8f714398c7a1a5a29adc2aed01857b41444ed53ec11863a3136ad37c8064c > docker exec $cid ipfs swarm peers /ip4/188.8.131.52/tcp/4001/ipfs/QmSoLpPVmHKQ4XTPdz8tjDFgdeRFkpV8JgYq8JVJ69RrZm /ip4/184.108.40.206/tcp/4001/ipfs/QmSoLSafTMBsPKadTEgaXctDQVcqN88CNLHXMkTNwMKPnu /ip4/220.127.116.11/tcp/4001/ipfs/QmSoLueR4xBeUbY9WZ9xGUUxunbKWcrNFTDAadQJmocnWm /ip4/18.104.22.168/tcp/4001/ipfs/QmSoLMeWqB7YGVLJN3pNLQpmmEk35v6wYtsMGLzSr5QBU3
> echo "hello from dockerized ipfs" >/tmp/ipfs-docker-staging/hello > docker exec $cid ipfs add /export/hello added QmcDge1SrsTBU8b9PBGTGYguNRnm84Kvg8axfGURxqZpR1 /export/hello > docker exec $cid ipfs cat /ipfs/QmSvCqazpuuib8qyRyddyFemLc2qmRukLLy8YfkdRPEXoQ hello there!
容器化的 IPFS 现在运行了一个网关，在
> curl http://localhost:8080/ipfs/QmcDge1SrsTBU8b9PBGTGYguNRnm84Kvg8axfGURxqZpR1 hello from dockerized ipfs
Kubernetes 1.9 已经来到, 我们可以基于Kubernetes构建 IPFS 节点的集群，可以存储任何数据，可以从任何 IPFS 集群中的节点上快速提取数据, 任何人都可以参与!