Last week, we got into problems when booting up our AKS cluster(we’ll shut the development cluster down every night to safe costs). Instead of green lights, our Neo4J database refused to run. In the logs, we noticed the following error message:
ERROR Invalid memory configuration - exceeds physical memory.
Let me share what caused this error.
Maybe you’ve read my article about resource limits in Kubernetes. There I talked about the fact that you can set resource limits at the container level.
What I didn’t mention in the article is that you can also configure default limits at the namespace level through limit ranges.
From the documentation:
A LimitRange provides constraints that can:
- Enforce minimum and maximum compute resources usage per Pod or Container in a namespace.
- Enforce minimum and maximum storage request per PersistentVolumeClaim in a namespace.
- Enforce a ratio between request and limit for a resource in a namespace.
- Set default request/limit for compute resources in a namespace and automatically inject them to Containers at runtime.
So if you don’t configure resource limits and/or requests at the container level, you can still set it at the namespace level.
This is exactly what we did, here are the limit ranges that are currently in place:
And it are these (default) limits that brought our Neo4J instance into trouble. Although enough memory was available in the cluster, the container was limited by default to only use 512MB which is unsufficient to run our Neo4J cluster. The solution was to change our Helm chart to assign more memory to the Neo4J pods.
When configuring resource limits, settings at the pod/container level always supersede settings at the namespace level.