Shulou(Shulou.com)05/31 Report--

Editor to share with you the example analysis of CERN evaluation using Rook to deploy Ceph, I believe that most people do not know much about it, so share this article for your reference. I hope you will learn a lot after reading this article. Let's go to know it together.

CERN Evaluation using Rook to deploy Ceph

CERN has been using Ceph since 2013. In addition to operating the largest Ceph cluster, it is also an active contributor to the Ceph community. CERN benefits from Ceph in many ways, including:

OpenStack mirroring and volumes (RBD)

HPC scratch space (CephFS)

Private file sharing similar to nfs (CephFS)

Object Storage (RGW) compatible with Amazon S3

CERN has to deal with pb-level data, so it is always looking for ways to simplify cloud-based deployment. It has been actively evaluating the container-based approach based on the Kubernetes infrastructure. One of the technologies that has attracted much attention recently is Rook; 's storage coordinator for cloud native environments. Although Ceph has been satisfying its storage use cases for many years, it will get better if its operational complexity can be reduced in some way. CERN currently uses a combination of kickstart and puppet to configure and manage Ceph deployments, which introduces some time constraints. Since the recent Rook Ceph integration is in the Beta phase, it is necessary to evaluate it to determine whether future Ceph clusters can be deployed using this tool. Specifically, you want to use Rook to improve the following areas:

Reduce deployment time for new clusters

Simplify upgrad

More flexible scale-out

Better fault tolerance

Reduce reliance on professional Ceph operators

My job is to deploy Ceph using Rook in both virtual and physical environments to see if the above improvements have been achieved. When the latest version of Rook is deployed, the installation process is straightforward; as long as you set up a Kubernetes environment, you can easily deploy it with two commands. An important aspect of the assessment is the ability to quickly add / remove OSDs, RGWs, and MDSs from a running cluster using ceph-mgr 's new orchestration CLI module. Since the docker image used by Rook v0.8 did not have the latest Ceph master at the time of my project, I had to build a custom docker image.

CERN's cloud infrastructure is driven by OpenStack, so the virtual environment consists of OpenStack's virtual machines, while the physical environment consists of bare metal servers managed by OpenStack Ironic, each with 48 disks with a capacity of 5.5 TB. My test platform included deploying S3 workloads to collect metrics such as the time to deploy the entire Ceph cluster, the time to add new services (RGW, OSD, MDS) to the cluster, the automation of Ceph upgrades, and the reliability of the entire cluster.

Rook didn't let me down on these indicators. The newly deployed Ceph cluster can be used in minutes. It takes less than 2 minutes to add a new OSD to the running cluster. The same is true for adding services such as RGW and MDS. Rook is based on Kubernetes, so node reliability and improved fault tolerance become first-class features. The choreographer CLI with the Rook backend makes it easy to manipulate deployed clusters, especially for those who are already familiar with Ceph CLI. However, the upgrade of Ceph is not yet automated, and there is coupling between Rook and Ceph versions, which reduces the flexibility of the overall deployment. These features are scheduled to be available in the upcoming release of Rook v0.9.

Because of the huge infrastructure of CERN, it makes sense to actively explore technology to make the operation faster and easier. In my opinion, Rook looks like a promising open source project with the goal of making storage more suitable for the cloud. If you can implement functions such as automatic upgrades and support for decoupled versions, it would be a good idea to use Rook and Kubernetes for deployment in future Ceph clusters. The results of this project will help CERN people decide whether they should use Rook and Kubernetes in future Ceph deployments.

The above is all the content of the article "CERN Evaluation of the sample Analysis of deploying Ceph using Rook". Thank you for reading! I believe we all have a certain understanding, hope to share the content to help you, if you want to learn more knowledge, welcome to follow the industry information channel!

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.