[Gluster-devel] kadalu - k8s storage with Gluster
amarts at gmail.com
Thu Sep 12 09:56:17 UTC 2019
Hi Gluster users,
I am not sure how many of you use Gluster for your k8s storage (or even
considering to use). I have some good news for you.
Last month, I along with Aravinda spoke at DevConf India, about project
kadalu. The code & README available @ https://github.com/kadalu/kadalu. We
are awaiting the talk's video to be uploaded, and once done I will share
the link here.
Wanted to share few highlights of the kadalu project with you all, and also
future scope of work.
- kadalu comes with *CSI driver*, so one can use this smoothly with k8s
- Has an *operator* which starts CSI drivers, and Gluster storage pod
- 2 commands to setup and get k8s storage working.
- kubectl create -f kadalu-operator.yml
- kubectl create -f kadalu-config.yml
- Native support for single disk use-case (ie, if your backend supports
High Availability, no need to use Gluster's replication), which I believe
is a good thing for people who already have some storage array which is
highly available, and for those companies which have their own storage
products, but doesn't have k8s expose.
- The above can be usecase can be used on a single AWS EBS volume, if
you want to save cost of Replica 3 (If you trust it to provide your
required SLA for it). Here, Single EBS volume would provide multiple k8s
- GlusterFS used is very light mode, ie, no 'glusterd', no LVM, or any
other layers. Only using glusterfs for filesystem, not management.
- Basic end to end testing is done using Travis CI/CD. [Need more help
to enhance it further].
More on this in our presentation @
Please note that this is a project which we started as a prototype for our
talk. To take it further, feedback, feature request, suggestions and
contributions are very important. Let me know if you are interested to
collaborate on this one.
Possible future work:
* Implement data backup features (possibly with geo-rep).
* Resize of Volume (Both backend Gluster volume, and PV volume).
* Consider implementing helm chart for operator.
* Scale testing, etc.
Limitations (for now)
* No 'migration'. One has to start fresh with kadalu.
* No Snapshot, No cloning.
* <Please report limitation you faced through github issues>
As of now, there are 2 deployment guides available @
Thanks & Regards,
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Gluster-devel