Skip to content

glusterfs kubernetes, maybe the simplest helm chart of glusterfs-centos4u1 application, with no complex ENV setup

Notifications You must be signed in to change notification settings

SobeyDigital/glusterfs-centos4u1-helm

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

6 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

simplest glusterfs helm chart

maybe the simplest helm chart of glusterfs-centos4u1 application, with no complex ENV setup

Very easy to setup

this application is the simplest way to deploy glusterfs container on kubernetes. It is based on gluster/gluster-kubernetes, but, there's no heketi need!

it's very fun that:

  1. you don't need install gluster-fuse or client on your node
  2. you dont't need heketi
  3. you can setup a cluster only one node (you don't need at least 2 or 3 node)
  4. if you use rancher, you need only fill node's ips and volume names of glusterfs.

How

because I want to build a contanierd glusterfs system on k8s, but I found that it's not easy for somebody because you need to setup very complex envirment, and you need at least 2 or 3 or more node (based on glusterfs system), and you must use heketi.

But, I only want to:

  1. simply pull and start the glusterfs container
  2. I don't want to use heketi because I want to manange my custom PV
  3. maybe I have only one Node

So... I try to make it this way:

a helm app with two SatefulSet k8s object

  • a) one is glusterfs centos4u1 (because PVC use 4.1 client to mount). And with a shell command that: -- 1) init some special things, mount self to the __vol_system__ volume -- 2) loop to find new node in __vol_system__ volume -- 3) probe new node in the cluster
  • b) another one is a boot pod named "shadowadmin", it's the key that: -- 1) init some special things -- 2) listen StatefulSet-0 glusterfs centos4u1 start -- 3) probe StatefulSet-0 glusterfs centos4u1 in gluster -- 4) create a __vol_system__ volume, and add bricks -- 5) create all voluem defined in the helm yaml, and add bricks -- 5) remove all brick of shadowadmin out of the gluster after that down, you got a "one node" glusterfs cluster, and because of (StatefulSet-0 glusterfs centos4u1)'s shell, you can setup one or more node.

note that, in github, I only push replica model

Need help

As you can see, this based on StatutfulSet object and "HostNetwork", that means you could just run one glusterfs pod in a node. The PV and PVC's behavior that it use EndPoint to find glusterfs's ip,if you dont't use hostNetwork, you'll found an error that says such like "xxx-0.xxx could not resovle", because the pv plugin now can not access the StatufulSet glusterfs pod's dns without namespace, and EndPoint only support selector or IP.

so if there is a new pv plugin, just tell me!

About

glusterfs kubernetes, maybe the simplest helm chart of glusterfs-centos4u1 application, with no complex ENV setup

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Shell 89.1%
  • Smarty 10.9%