Kubernetes monitoring is a crucial part. I started reading about it and I was confused with lot of terms used in various blogs posts and github official documentation. In this article I am trying to demystify all of the commonly used terms which are overwhelming for over beginner.

In one of previous post we have seen the hyperconverged implementation of ceph using ROOK on kubernetes. In this article, I am showing the integration of existing ceph cluster with kubernetes, in other words, using existing ceph as backend for POD persisten storage.

Custom resource defintions which are created by Rook operator.

Running a stateful application in POD is not an easy task. I worked on ceph for brief period of time. I thought of using an existing ceph setup to provide persistent storage for kubernetes. For this exercise, created single node ceph cluster. I was facing some issues in integrating existing ceph setup with minikube kubernetes. While I was searching for this issue, I came across this interesting project ROOK which is used to create ceph cluster on top of kubernetes. I really like that idea, in case of on-premise kubernetes deployment where servers are having lot of disks, and these disks can be used to create a distributed ceph cluster using ROOK. It’s not only limited to ceph only other supported storage types are “CockroachDB” and “minio” as per project github page. Okay enought talking, let’s start with some practical work.

Helm is the package manager for kubernetes, similar to dnf or yum in linux. It’s very to start with helm, if you have followed my previous articles, you may already know that I am using minikube for kubernetes learning which is running on Mac using Oracle virtualbox provider. Helm is based on client/server architecture, it has two components helm itself which is command to talk to kube-api server in my case helm is installed on Mac and tiller which installed on kubernetes nodes i.e minikube.

In this article, I am covering the kubernetes storage options for stateful containers. To persist the container data we need to attach the volume with container. Kubernetes has evolved a lot to provide the various storage options to create the volumes required by POD.

This is just for my own reference for seeing the options with which by default services are getting started in minikube.

In the previous article, we have seen some default POD and services which are started by default. In this article, I am going to spin up a POD and check the stuff related to POD.

Logging is very crucial topic for containers. By default like docker, kubernetes use the json log driver which doesn’t support the multi-line log which means that it will not be able to dump the call traces in log file. Unfortunately K8 doesn’t provide the option to change the logging driver for docker but we can do that while starting the container using docker run command. There is open feature request [1] for this.

Getting the following error while trying to start the kubernetes POD.