If your Kubernetes Azure AKS cluster is old enough, especially in development environments, it will definitely run out of disk space, because AKS will store all container images (and in DEV environments they're changing quite often) on the disk.

I wasn't able to find, how often AKS engine is executing some clean operation (but it does, absolutely). However, in my case, space usage reached those already critical 80% and continued to grow. Here's how can run cleaning manually.

When you create Azure Kubernetes Service cluster (AKS), you can specify Log Analytics resource for logging and monitoring of Kubernetes resources: pods, services, configmaps, etc. This thing is actually installing Log Analytics agent inside your cluster (there will be several pods and related resources like configmaps), which then will send stdout and stderroutputs from pods to Log Analytics.

Unfortunately, the schema of these logs inside Log Analytics is quite complex, mostly because that the log record is coming from container, but not the pod, so they have multiple tables, which we must connect to get human-readable logging data.

My customer require that we have supported versions of software installed, and I'm forced to upgrade Kubernetes as well according to the support policy of Kubernetes releases in Azure.

Last Kubernetes 1.24+ upgrade was not very clean - Nginx-ingress controller failed, and all our applications stopped responding.