Simon Krenger

One of the newer features in Kubernetes (1.30 and later) is the Kubelet Checkpoint API. This new API allows users to create a stateful copy of a running container, a functionality which is often used for forensics or for debugging.

In Kubernetes installations where this feature is enabled, a checkpoint can be created by accessing the respective Kubelet API via curl or similar. In the following example I am also using the Kubernetes API /proxy endpoint (the same can also be done on the Node locally via localhost:10250/checkpoint/...):

$ curl -k -X POST --header "Authorization: Bearer $TOKEN" "$KUBERNETES_API_URL/api/v1/nodes/$NODE_NAME/proxy/checkpoint/$NAMESPACE_NAME/$POD_NAME/$CONTAINER_NAME"
{"items":["/var/lib/kubelet/checkpoints/checkpoint-fedora-74d79dd7f4-csrmg_skrenger-container-2024-12-12T12:56:19Z.tar"]}

Well, so I tried installing a new ARM-based OpenShift Container Platform cluster on AWS. To prepare, I created an install-config.yaml file and changed the controlPlane.architecture and the compute.architecture field to “arm64” and then launched the installer. That did not work, it still complains about the architecture:

$ ./openshift-install create cluster --dir=.
INFO Credentials loaded from the "default" profile in file "/home/simon/.aws/credentials" 
INFO Consuming Install Config from target directory 
INFO Creating infrastructure resources...         
INFO Waiting up to 20m0s (until 11:07AM) for the Kubernetes API at https://api.skrenger-arm.lab.example.com:6443... 
INFO Pulling VM console logs                      
INFO Pulling debug logs from the bootstrap machine 
ERROR Attempted to gather ClusterOperator status after installation failure: listing ClusterOperator objects: Get "https://api.skrenger-arm.lab.example.com:6443/apis/config.openshift.io/v1/clusteroperators": dial tcp 3.64.25.143:6443: connect: connection refused 
ERROR Bootstrap failed to complete: Get "https://api.skrenger-arm.lab.example.com:6443/version": dial tcp 3.68.144.150:6443: connect: connection refused 
ERROR Failed waiting for Kubernetes API. This error usually happens when there is a problem on the bootstrap host that prevents creating a temporary control plane. 
ERROR The bootstrap machine failed to download the release image 
INFO Pulling quay.io/openshift-release-dev/ocp-release@sha256:9ffb17b909a4fdef5324ba45ec6dd282985dd49d25b933ea401873183ef20bf8... 
INFO cfce1ab124f59e93a0f67d7e85283d524ddfd73a27d0535319d69d1dce746488 
INFO ERROR: release image arch amd64 does not match host arch arm64 
INFO Bootstrap gather logs captured here "/home/simon/Downloads/arm/log-bundle-20221124110737.tar.gz" 

In OpenShift Container Platform (OCP) 4, most of the functionality is controlled by Operators. To see the currently installed Operators and also their status, use the following command:

$ oc get clusteroperators
NAME                                       VERSION   AVAILABLE   PROGRESSING   DEGRADED   SINCE
authentication                             4.6.4     True        False         False      12m
cloud-credential                           4.6.4     True        False         False      38m
cluster-autoscaler                         4.6.4     True        False         False      32m
config-operator                            4.6.4     True        False         False      33m
console                                    4.6.4     True        False         False      21m
csi-snapshot-controller                    4.6.4     True        False         False      27m
dns                                        4.6.4     True        False         False      31m
etcd                                       4.6.4     True        False         False      32m
image-registry                             4.6.4     True        False         False      25m
ingress                                    4.6.4     True        False         False      24m
insights                                   4.6.4     True        False         False      33m
kube-apiserver                             4.6.4     True        False         False      30m
kube-controller-manager                    4.6.4     True        False         False      31m
kube-scheduler                             4.6.4     True        False         False      31m
kube-storage-version-migrator              4.6.4     True        False         False      24m
machine-api                                4.6.4     True        False         False      27m
machine-approver                           4.6.4     True        False         False      32m
machine-config                             4.6.4     True        False         False      32m
marketplace                                4.6.4     True        False         False      32m
monitoring                                 4.6.4     True        False         False      23m
network                                    4.6.4     True        False         False      33m
node-tuning                                4.6.4     True        False         False      33m
openshift-apiserver                        4.6.4     True        False         False      27m
openshift-controller-manager               4.6.4     True        False         False      24m
openshift-samples                          4.6.4     True        False         False      26m
operator-lifecycle-manager                 4.6.4     True        False         False      32m
operator-lifecycle-manager-catalog         4.6.4     True        False         False      32m
operator-lifecycle-manager-packageserver   4.6.4     True        False         False      27m
service-ca                                 4.6.4     True        False         False      33m
storage                                    4.6.4     True        False         False      32m

You can find the description of the default Operators in the documentation.

This will only list the Red Hat Operators that are installed as part of the cluster. These are all controlled by the ClusterVersionOperator, which is the “Master-Operator” of the cluster controlling all others.

If you want to list all Operators that were installed via the Operator Lifecycle Manager (OLM), you can use the following command:

$ oc get subscriptions --all-namespaces

Getting training and exams done in 2020 has been challenging. After reaching my RHCE mid-February, I am now proud to say that I achieved my Red Hat Certified Architect in Infrastructure certification less than 9 months later.

To reach my RHCA, I took the following Red Hat exams. As you can see, it is OpenShift and Ansible all the way down:

• EX180 Red Hat Certified Specialist in Containers and Kubernetes
• EX280 Red Hat Certified Specialist in OpenShift Administration
• EX288 Red Hat Certified Specialist in OpenShift Application Development
• EX407 Red Hat Certified Specialist in Ansible Automation
• EX447 Red Hat Certified Specialist in Ansible Best Practices

Of course, the journey does not end here as there are quite a few interesting topics still to learn!

With OpenShift 4, Red Hat introduced Red Hat Enterprise Linux CoreOS. It is a very minimalist operating system, focused on running container workload.

This new minimalism comes with some challenges. There are no more RPM packages and most of the tools we know and love are missing! Luckily, there is the Red Hat supplied toolbox container that contains all the necessary tools and is nicely integrated.

So to start the toolbox, use oc debug node/<nodename>. This will start a privileged container on the node you specify, mount the host file system on /host and drop you into a shell:

$ oc debug node/worker-0.lab.openshift.krenger.ch
Starting pod/worker-0labopenshiftkrengerch-debug ...
To use host binaries, run `chroot /host`
If you don't see a command prompt, try pressing enter.
sh-4.2# chroot /host
sh-4.4# toolbox
Container started successfully. To exit, type 'exit'.
sh-4.2#

Now we are running in the toolbox container on our CoreOS host with all the tools we know at our disposal, for example sosreport:

sh-4.2# sosreport

Running sosreport will generate a sosreport in /host/var/tmp/, which means it will be accessible in /var/tmp/ on the CoreOS host itself.

For OpenShift 4, the upgrade paths are kept in the cincinnati-graph-data repository as YAML files and then exposed via an API.

There is a Red Hat Solution describing how this data can be queried via api.openshift.com and how you can use this data in your automation:

$ curl -sH 'Accept:application/json' 'https://api.openshift.com/api/upgrades_info/v1/graph?channel=fast-4.2&arch=amd64' | jq .

While this data is quite helpful for automation (the Solution also describes helpful queries), it is not very nice to look at the raw data. If you are looking for a graphical presentation of that data, you should check out this wonderful website that is maintained by a Red Hat colleague with hourly generated data: www.ocp-upgrade.net

So here is another one from the trenches.

More than once one of our OpenShift Container Platform customers approached us and said something along the lines of: “Help, I cannot see the X-Forwarded-For header in my application, our OpenShift Router is probably configured incorrectly!”.

In such cases, it is often a good idea to check what is really being forwarded to the Pods in the cluster. For this, I typically use my simonkrenger/echoenv container to print the headers received by the application. In many cases, it turns out that the application affected is a Spring Boot application and the header is passed correctly to the Pod itself. But the Spring Boot application does not show the header anyway.

We have observed a behaviour of Spring Boot that leads to the X-Forwarded-For header not being passed to the application, as it is consumed by Spring Boot. In the application.properties of a Spring Boot application, the following setting controls this:

server.use-forward-headers: true

This configuration leads to the header being consumed by Spring Boot and the header not being available in the application. See also the relevant sections in Spring documentation. Good to know.

Kubernetes uses etcd as the persistent store for API data. As etcd is a distributed key-value store, we can also use command line tools to query this store. The examples in this post are for OpenShift 3.x.

Apart from just using get, there is also the possibility to perform the following actions on certain keys:

• put to write to a key – unless you know what you are doing, don’t touch the Kubernetes data in etcd, as this will manifest in very strange Kubernetes behaviour.
• del to delete a key – also, this may break your Kubernetes cluster by introducing inconsistencies.
• watch to keep a watch on an object. This is very helpful to track changes on a certain object.

The get action is probably the most helpful functionality for in-depth API debugging directly within etcd.

For editing YAML, be it for OpenShift / Kubernetes or Ansible, having your editor set up right can help to avoid common mistakes. So here is the minimalistic config in my ~/.vimrc to make working with YAML files easier. I am sure there are even more plugins or settings available, but this minimal set of commands works fine for me:

set ts=2
set sts=2
set sw=2
set expandtab

syntax on
filetype indent plugin on

set ruler

Some time ago, I had a curious case of very slow DNS resolution in a container on OpenShift. The symptoms were as follows:

• In the PHP application in the container, DNS resolution was very slow with a 5 second delay before the lookup was resolved
• In the container itself, DNS resolution for curl was very slow, with a 5 second timeout before the lookup was resolved
• However, using dig in the container itself, DNS resolution was instant
• Also, on the worker node, the DNS resolution was instant (using both dig and curl)

TL;DR: Since glibc 2.10, glibc performs IPv4 and IPv6 lookups in parallel. When IPv6 fails, there is a 5 second timeout in many cases before the lookup is returned. Disable IPv6 DNS lookups by setting “single-request” in “resolv.conf” or disable the IPv6 stack completely.