Part 1

Introduction to Debugging

Kubernetes is a "self-healing" system, and we'll get back to what Kubernetes consists of and how it actually works in part 5. But at this stage "self-healing" is an excellent concept: Often you (the maintainer or developer) don't have to do anything in case something goes wrong with a pod or a container.

Sometimes you need to interfere, or you might have problems with your own configuration. As you are trying to find bugs in your configuration start by eliminating all possibilities one by one. The key is to be systematic and to question everything. Here are the preliminary tools to solve problems.

The first is kubectl describe which can tell you most of everything you need to know about any resource.

The second is kubectl logs with which you can follow the logs of your possibly broken software.

The third is kubectl delete which will simply delete the resource and in some cases, like with pods in deployment, a new one will be automatically released.

Finally, we have the overarching tool Lens "The Kubernetes IDE". Which you should start using right now to familiarize yourself with the usage.

During exercises, you also have our Telegram group available (which you joined in part 0).

Let's test these tools and experiment using Lens. You will likely face a real debugging challenge during the exercises and there is another preplanned one in part 5 when we have a larger set of moving parts available to us.

Let's deploy the application and see what's going on.

$ kubectl apply -f https://raw.githubusercontent.com/kubernetes-hy/material-example/master/app1/manifests/deployment.yaml
  deployment.apps/hashgenerator-dep created

$ kubectl describe deployment hashgenerator-dep
  Name:                   hashgenerator-dep
  Namespace:              default
  CreationTimestamp:      Wed, 16 Sep 2020 16:17:39 +0300
  Labels:                 <none>
  Annotations:            deployment.kubernetes.io/revision: 1
  Selector:               app=hashgenerator
  Replicas:               1 desired | 1 updated | 1 total | 1 available | 0 unavailable
  StrategyType:           RollingUpdate
  MinReadySeconds:        0
  RollingUpdateStrategy:  25% max unavailable, 25% max surge
  Pod Template:
    Labels:  app=hashgenerator
    Containers:
     hashgenerator:
      Image:        jakousa/dwk-app1:78031863af07c4c4cc3c96d07af68e8ce6e3afba
      Port:         <none>
      Host Port:    <none>
      Environment:  <none>
      Mounts:       <none>
    Volumes:        <none>
  Conditions:
    Type           Status  Reason
    ----           ------  ------
    Available      True    MinimumReplicasAvailable
    Progressing    True    NewReplicaSetAvailable
  OldReplicaSets:  <none>
  NewReplicaSet:   hashgenerator-dep-75bdcc94c (1/1 replicas created)
  Events:
    Type    Reason             Age    From                   Message
    ----    ------             ----   ----                   -------
    Normal  ScalingReplicaSet  8m39s  deployment-controller  Scaled up replica set hashgenerator-dep-75bdcc94c to 1

There's a lot of information we are not ready to evaluate yet. But take a moment to read through everything. There're at least a few key information pieces we know, mostly because we defined them earlier in the yaml. The events are often the place to look for errors.

The command describe can be used for other resources as well. Let's see the pod next:

$ kubectl describe pod hashgenerator-dep-75bdcc94c-whwsm
  ...
  Events:
    Type    Reason     Age   From                              Message
    ----    ------     ----  ----                              -------
    Normal  Scheduled  15m   default-scheduler                 Successfully assigned default/hashgenerator-dep-75bdcc94c-whwsm to k3d-k3s-default-agent-0
    Normal  Pulling    15m   kubelet, k3d-k3s-default-agent-0  Pulling image "jakousa/dwk-app1:78031863af07c4c4cc3c96d07af68e8ce6e3afba"
    Normal  Pulled     15m   kubelet, k3d-k3s-default-agent-0  Successfully pulled image "jakousa/dwk-app1:78031863af07c4c4cc3c96d07af68e8ce6e3afba"
    Normal  Created    15m   kubelet, k3d-k3s-default-agent-0  Created container hashgenerator
    Normal  Started    15m   kubelet, k3d-k3s-default-agent-0  Started container hashgenerator

There's again a lot of information but let's focus on the events this time. Here we can see everything that happened. Scheduler put the pod to the node with the name "k3d-k3s-default-agent-0" successfully pulled the image and started the container. Everything is working as intended, excellent. The application is running.

Next, let's check that the application is actually doing what it should by reading the logs.

$ kubectl logs hashgenerator-dep-75bdcc94c-whwsm
  jst944
  3c2xas
  s6ufaj
  cq7ka6

Everything seems to be in order. However, wouldn't it be great if there was a dashboard to see everything going on? Let's see what the Lens can do.

First, you'll need to add the cluster to Lens. If the config is not available in the dropdown you can get the kubeconfig for custom with kubectl config view --minify --raw. After you've added the cluster open Workloads/Overview tab. A view similar to the following should open up

lens during deploy

At the bottom, we can see every event, and at the top, we can see the status of different resources in our cluster. Try deleting and reapplying the deployment and you should see events in the dashboard. This is the same output that you would see from kubectl get events

Next, let's navigate to the tab Workloads/Pods and click our pod with the name "hashgenerator-dep-...".

lens pod

The view shows us the same information as was in the description. But the GUI offers us actions as well. The three numbered in the top right corner are:

  1. Open terminal into a container in the pod
  2. Show logs
  3. Delete the resource

In addition, at the bottom, you can open a terminal with the correct context.

"The best feature in my opinion is that when I do kubectl get pod in the terminal, the dashboard you are looking at is always in the right context. Additionally, I don't need to worry about working with stale information because everything is real-time." - Matti Paksula

:
Loading...
:
Loading...

Login to view the exercise

You have reached the end of this section! Continue to the next section: