There are a lot of use-cases where a dedicated isolated k8s cluster would be the natural choice, BUT:
its to complex to manage a lot of small (very similar) physical clusters
all you got is a restricted access: 1 single namespace in a shared cluster, with limited RBAC (no cluster wide rights)
you are using a cloud provider's k8s and can't change some config.
Enter the Matrix, I mean Virtual Clusters. You can have your own apiserver+etcd with full control, but workload is still running on the host/physical/outer cluster. So conceptually it's not k8s-in-k8s (like docker-in-docker)
Additional use-cases:
you want to use a different k8s version than the hosting cluster.
you need multiple versions of the same app (multiple istio versions)
you want to run/test admission controllers, but afraid of breaking some other part of the cluster
you want to test a newer version of k8s, while your cloud provider is behind
I will demonstrate how easy it is to setup and use the open-source vcluster project.
Lajos co-founded SequenceIQ in early 2014. The small startup was focused on letting Hadoop run in docker containers in a cloud-agnostic way. The ambitious plan was getting fulfilled after a bit more than 1 year, when Hortonworks acquired them in 2015 April.
In 2017 BanzaiCloud was born to continue the cloud-native journey, to make running and managing containerised workload easier.
In 2018 focus has been shifted towards consultation and delivering trainings/workshops. Notable conferences include: GotoBerlin, GotoCopenhagen, CraftConf.
He is the organiser of the Docker Budapest Meetup and co-organises the Kubernetes Meetup.
