Rook is a distributed storage system designed for cloud native applications. It exposes file, block, and object storage on top of shared resource pools. Rook has minimal dependencies and can be deployed in dedicated storage clusters or converged clusters. It's self-managing, self-protecting, self-healing, and is designed to just work without teams of engineers managing it. It scales from a single node, to multi-PB clusters spread geographically. It's based on the Ceph project with over 10 years of production deployments in some of the largest storage clusters in the world.
Rook is in alpha state. We're just getting started. Not all planned features are complete. The API and other user-facing objects are subject to change. Backward-compability is not supported for this release. See our Roadmap and Issues. Please help us by Contributing to the project.
Here's the quickest way to get going with Rook.
On a modern Linux host run the following:
-
Download the latest binaries
$ wget https://github.com/rook/rook/releases/download/v0.1.0/rook-v0.1.0-linux-amd64.tar.gz $ tar xvf rook-v0.1.0-linux-amd64.tar.gz
-
Start a one node Rook cluster
$ ./rookd --data-dir /tmp/rook-test
-
Now in a different shell (in the same path) create a new volume image (10MB)
$ ./rook block create --name test --size 10485760 -
Mount the block volume and format it
sudo ./rook block mount --name test --path /tmp/rook-volume sudo chown $USER:$USER /tmp/rook-volume
-
Write and read a file
echo "Hello Rook!" > /tmp/rook-volume/hello cat /tmp/rook-volume/hello
-
Cleanup
sudo ./rook block unmount --path /tmp/rook-volume
To run a Kubernetes cluster with Rook for persistent storage go here
Rook is also easy to run on CoreOS either directly on the host or via rkt.
cd demo/vagrant
vagrant up
See Building in the wiki for more details.
A rook cluster is made up of one or more nodes each running the Rook daemon rookd. Containers and Pods can
mount block devices and filesystems exposed by the cluster, or can use S3/Swift API for object storage. There is
also a REST API exposed by rookd as well as a command line tool called rook.
The Rook daemon rookd is a single binary that is self-contained and has all that is needed to bootstrap, scale
and manage a storage cluster. rookd is typically compiled into a single static binary (just like most golang
binaries) or a dynamic binary that takes a dependency on mostly libc. It can run in minimal containers, alongside a
hypervisor, or directly on the host on most Linux distributions.
rookd uses an embedded version of Ceph for storing all data -- there are no changes to the data path. An embedded version
of Ceph was created specifically for Rook scenarios and has been pushed upstream. Rook does not attempt to maintain full fidelity
with Ceph, for example, most of the Ceph concepts like OSDs, MONs, placement groups, etc. are hidden. Instead Rook creates
a much simplified UX for admins that is in terms of physical resources, pools, volumes, filesystems, and buckets.
rookd embeds Etcd to store configuration and coordinate cluster-wide management operations. rookd will automatically
bootstrap Etcd, manage it, and scale it as the cluster grows. It's also possible to use an external Etcd instead of the embedded one
if needed.
Rook and etcd are implemented in golang. Ceph is implemented in C++ where the data path is highly optimized. We believe this combination offers the best of both worlds.
See Design wiki for more details.
We welcome contributions. See Contributing to get started.
For filing bugs, suggesting improvements, or requesting new features, help us out by opening an issue.
Please use the following to reach members of the community:
Rook and Etcd are under the Apache 2.0 license. Ceph is mostly under the LGPL 2.0 license. Some portions of the code are under different licenses. The appropriate license information can be found in the headers of the source files.