About the project and its status
Longhorn was introduced in April 2017 and is called distributed block storage for Kubernetes clusters. Among its main features are fault tolerance (there is no single point of failure) and focus on application in enterprise, incremental snapshots, backups to additional storage (NFS or compatible with S3), automated updates without downtime, easy to use GUI.
Longhorn Storage Organization Diagram
Longhorn Panel Web Interface
The source code for the engine for this solution ( longhorn-engine ) is written in Go and Python. For its work, it requires Docker v1.13 + and Kubernetes v1.8 + (the recommended version is 1.12+). Of the file systems on the host OS where the data will be stored, ext4 and xfs are supported ( extent support is required).
The current level of readiness of Longhorn is the alpha version , the latest release of the project is 0.6.2 (for more details, see the main repositories ). The number of contributors to the project’s code base is estimated at several dozens (the most - 17 - at the longhorn-engine repository), and the number of stars on GitHub has not yet reached 1000.
Installing Longhorn into an existing Kubernetes cluster is done with a single
kubectl apply
or using Helm charts. Instructions can be found here . After installation in the cluster, the appropriate driver will become available: CSI or Flexvolume, depending on the K8s version, for connecting persistent volumes based on Longhorn.
On the Internet you can find comparisons of Longhorn with other repositories for Kubernetes:
- Comparingforgeeks features comparison : OpenEBS, Portworx, Rook, StorageOS, Longhorn, GlusterFS (September'19);
- short descriptions + performance comparison from Vito Botta : OpenEBS, Rook (Ceph), Longhorn, StorageOS, Robin, Portworx, Linstor (August'19);
- simple performance comparison on Reddit : Longhorn, Rook (Ceph), StorageOS (April'19).
Longhorn showed average performance results compared to other storages according to the results of testing conducted by Vito Botta (for details, see the link above)
Longhorn at CNCF
In a CNCF organization that supports infrastructure Open Source solutions for modern cloud native applications, Longhorn has expanded its sandbox with 20 other projects such as SPIFFE , Telepresence , OpenMetrics, OpenTelemetry, and KubeVirt.
One of the criteria for moving from the sandbox to the next level of maturity of CNCF projects - to the incubating - is the proof that at least three independent organizations of an “adequate scale” work with the solution in production. This status already has, for example, Rook.
- You can read more about the criteria adopted by CNCF in this document .
- A complete list of projects that have reached the greatest maturity (graduated and incubating levels) is presented here .
NB : By the way, recently (October 31), the graduated list was replenished with the Jaeger distributed tracing solution, which has already become quite popular in the cloud native world.
Sheng Liang, CEO at Rancher Labs, in an interview with The New Stack:
“I think it will benefit everyone (win-win). This is great news for the community because Longhorn is just fantastic technology; of which CNCF is needed to further popularize technologies like Kubernetes. Therefore, good for CNCF. It’s also great for us at Rancher, because we are completely absorbed in working on K3S [the lightweight Kubernetes distribution] and Rancher. Rancher will still not become a storage company. We are engaged in the monetization and creation of the Kubernetes enterprise platform as such. By making Longhorn part of the CNCF, we can attract more people to use the project. ”
PS
Read also in our blog: