Solved

OKD (Origin Kubernetes Distribution) + Kasten


Userlevel 7
Badge +6

Hi! Please, for testing and lab purposes, we need to deploy 02 x local K8s clusters using OKD (Origin Kubernetes Distribution) - the community distribution of OpenShift - and perform integration with Kasten. Later, we can migrate to the OpenShift production environment. Has any community member completed this deployment and integration (OKD +K10)? If so, could you share some guidelines/recommendations? We want to host the solution, at this time, on vSphere-based VMs. Thanks for everyone's help and attention! Thanks so much, Luiz Eduardo Serrano.

icon

Best answer by Geoff Burke 18 July 2023, 02:08

View original

7 comments

Userlevel 7
Badge +22

I tried but did not have any success. I tried on CRC and the new Openshift local as well, same result. The problem I believe is that Openshift is so heavy with its operators that you can’t get something like Kasten going without having to go in and tinker/disable other things. If I remember correctly you can’t even get the metrics server running on OKD due to limitations.

 

I might try again on Openshift local for fun to see but doubt I will get it going. 

Userlevel 7
Badge +6

Thank you, @Geoff Burke ! Do you believe the performance limitations were due to the use of virtualized servers/resource contention? We are considering using servers with a good amount of phy cores and RAM to host these VMs. Do you see any point of attention in this strategy? It would be great to hear your opinion.

Userlevel 7
Badge +22

I think it is an “on purpose” design thing. They don’t want people using these in production as I understand it. It is single node only so that alone will impede it. If you take Rancher Desktop for example it is a single node k3s but really light weight. If you loaded it with all the operators in Openshift I am sure it would slow down to a halt as well :) 

Userlevel 7
Badge +6

Hi, @Geoff Burke ! I'll consult the OS sizing guides and talk to the Red Hat team for recommendations. I will post here any news. Thanks for your help and comments!

Userlevel 7
Badge +22

It will be interesting to see what they say. I am guessing they will state that Openshift-Local and the previous versions Minishift, CRC, are meant for developers testing only and not production like workloads. I suspect they will point you here to “try Openshift” page. I believe this does let you setup temporarily a normal Openshift cluster and could then deploy it and test with Kasten.

https://www.redhat.com/en/technologies/cloud-computing/openshift/try-it?sc_cid=7013a000003ScmnAAC&gclid=Cj0KCQjw8NilBhDOARIsAHzpbLC9AeczR3VTdiojRbovwZi3JyRoTpLEBZsoBS9KtvgA0uf0T_s9kN4aAp8mEALw_wcB&gclsrc=aw.ds

I did once get Kasten installed from the Openshift Market place on Openshift-Local but the pods kept dying and in short would not work.

 

Either way get back to us on what they say. Maybe state that you just need to test kasten so perhaps they could advise on what other operators might be able to be turned off to free up resources.

 

cheers

 

Userlevel 7
Badge +6

Hey, sorry late to this one and intrigued to hear more. 

 

OpenShift does provide the ability to spin up clusters for I think 30 or 60 days which might be enough for your POC? This will unlock the full potential of what you are then going to see later on down the line vs using OKD which seems in line with as Geoff has said to K3s and Rancher Desktop, as well as Minikube and KinD. 

 

Userlevel 7
Badge +6

Hey, sorry late to this one and intrigued to hear more. 

 

OpenShift does provide the ability to spin up clusters for I think 30 or 60 days which might be enough for your POC? This will unlock the full potential of what you are then going to see later on down the line vs using OKD which seems in line with as Geoff has said to K3s and Rancher Desktop, as well as Minikube and KinD. 

 

Hello Michael! Thank you so much for the comments and guidance! Surely we will return to this theme. I am preparing an infrastructure to deploy the OS cluster for labs and K10 demonstrations. Let's evolve together! :-) Sincerely, Luiz Serrano. 

Comment