GlusterCS Deploy tool
A tool to deploy GlusterCS in a existing Kubernetes cluster.
Install the tool in the Kubernetes master node using,
sudo pip3 install kubectl-gluster
Note: Python 2 support is not yet available
Usage:
kubectl gluster deploy <config-yaml>
Example config file,
---
namespace: gcs
cluster-size: 3
nodes:
- address: kube1
devices: ["/dev/vdc", "/dev/vdd", "/dev/vde"]
- address: kube2
devices: ["/dev/vdc", "/dev/vdd", "/dev/vde"]
- address: kube3
devices: ["/dev/vdc", "/dev/vdd", "/dev/vde"]Where,
namespace - Cluster namespace, useful when managing multiple
clusters
cluster-size - Number of nodes in Gluster cluster. Currently only 3
nodes are supported.
nodes - Details of nodes where Gluster server pods needs to be
deployed.
address - Address of node as listed in kubectl get nodes(Use
`kubectl get nodes` to get the address)
devices - Raw devices which are required to auto provision
Gluster bricks during Volume create
How it works?
kubectl gluster deploy does the following tasks based on the provided
configuration.
-
Generate the Cluster ID.
-
GCS Namespace setup
-
Etcd Operator setup and Etcd cluster setup
-
CSI driver setup
-
Monitoring Setup
Is this a Operator?
No. But this tool can be called from an Operator. It is easy! generate the config when CRD changes and call this tool from Operator pod.
How is it different from gcs/deploy?
This project also uses manifest files from the gcs project, but executes Kube commands directly instead of using ansible playbook.
TODO
-
Currently using
kubectlcommand, instead use kubernetes client library. -
Improve error messages and usability.
-
Use kubernetes watches instead of polling for pods status
-
Python 2 support
-
Support for RPM/Deb packages
