Category: kubernetes

How to Return Custom Response Headers with Nginx Ingress in Kubernetes This method is useful when you have multiple Ingresses and want to confirm which one is responding.

If you encounter the error "no matches for kind 'CronJob' in version 'batch/v1beta1'" when registering a Kubernetes CronJob, you can resolve it by changing batch/v1beta1 to batch/v1.

I was using MicroK8s and found that the Pods were unable to communicate with the external network. Reinstalling the DNS plugin (reconfiguring the DNS server) in MicroK8s resolved the issue.

This article provides key points to check the configuration in cases where you are using the Nginx Ingress Controller with Kubernetes and Uwsgi for the application server within the Pod, to prevent slow response services from timing out. If these configurations are not extended, it could result in 502 or 504 errors.

I attempted to retrieve the source global IP address from HTTP requests through Kubernetes Ingress, but I ended up obtaining the IP address of Kubernetes' local network, which did not work as intended. After switching Kubernetes to MicroK8s and using its built-in Ingress controller, I was able to successfully retrieve the global remote IP address. This is a note on that experience.

When using MicroK8s on a server, a separate image repository for ctr is created apart from the server's Docker image repository. Since MicroK8s uses this ctr, Docker-loaded images cannot be used directly. It is necessary to import the images using `microk8s ctr image import`.

In the latest version of Kubernetes (around 1.25?), "apiVersion: extensions/v1beta1" is no longer supported. If you have written Ingress configurations using "extensions/v1beta1", applying them with "kubectl apply" will result in the error: error: unable to recognize "ingress.yml": no matches for kind "Ingress" in version "extensions/v1beta1" To resolve this issue, you need to update the "apiVersion" to "networking.k8s.io/v1".

How to write a simple host-to-IP address translation table, equivalent to the /etc/hosts file in an OS, in a Kubernetes manifest file.

Sure, I'd be happy to help. Could you please provide the Japanese blog article summary that you would like translated into English?

When operating a container with the tag set to "latest" and the imagePullPolicy set to "Always" within a Pod's Deployment in Kubernetes, you need to run `kubectl delete pod` to reflect the latest code. By staggering the deletion of pods, you can avoid downtime.

You can achieve this by writing the installation script in the args section of deployment.yml.

After using a large Kubernetes cluster within the company, I overlooked something when switching to a small EKS cluster for testing. When I checked the deployment to try to start the alb-ingress-controller, I found that the READY status was 0.

Pod exits with the error: "The node was low on resource: ephemeral-storage." Storage usage of the node is at 85% There is an unusually heavy process running in Kubernetes, causing the load average to be abnormally high (like 100) In this case, it is a storage shortage.

Here's a summary of the Japanese blog article in English: --- The article discusses installing k3s Kubernetes on a Raspberry Pi running Raspbian. It mentions that k3s will be installed in Docker mode. According to the Docker website, it is recommended to use the convenience script for Raspbian, so the article follows this recommendation.

When uploading a file to a web application running under Kubernetes Ingress, you might encounter a 413 Payload Too Large error. In such cases, you need to modify the Ingress settings.