Skip to content

Latest commit

 

History

History
213 lines (134 loc) · 8.79 KB

README.md

File metadata and controls

213 lines (134 loc) · 8.79 KB

kubernetes-localdev

Work In Progress (WIP)

For those who wish to deepen their knowledge, I'll be adding some Hands-on challenges.

Table Of Contents

1 - Earn Makefile

Scenario

Willy Wonka, a junior DevOps enginner, protected his Makefile, that is located at ch1-earn-makefile/earn-me.

Challenge

Find a way to read Willy's Makefile so that it will make sense.

The prize is quite lovely, and Willy wrote a Makefile which is related to this repository.

Hint

Expand/Collapse

How would you decode a Kubernetes Secret?

Solution

Expand/Collapse

The content of ch1-earn-makefile/earn-me seems like it's encoded with base64. How can I assume that? A "long" string that ends with == is usually (not always) the base64 representation of a normal string. When I see those kinds of strings, I attempt to decode them with the built-in command base64.

Here's how you can decode the Makefile using a one-liner Bash script.

base64 --decode challenges/ch1-earn-makefile/earn-me > Makefile
# A new file was created in $PWD
cat Makefile

Makefile (Short output)

.EXPORT_ALL_VARIABLES:
CATS_APP_NAME ?= baby
...
clean-minikube:
    minikube delete --purge --all

Use make help to check for available commands.

help                 Available make commands
start-minikube       Start minikube with docker driver and k8s version v1.20.2
check-versions       Check versions of relevant CLIs
helm-add-nginx-repo  Add and update the Helm repo ingress-nginx
run-cats             Run docker-cats locally

Conclusion

Upon completing the challenge - Expand/Collapse

Encoding is nothing more than a different representation of the same thing. Just as 11 in binary is 3 in decimal, you need to know the base, and that's it; there's no protection at all.

Willy should've used an Encryption mechanism, such as password, private+public keys, etc., to protect his Makefile.

2 - Change The Image

Scenario

As part of testing docker-cats, Willy wants to change the image of baby.jpg (even though it's adorable). The change should take place immediately, with minimum effort.

Challenge

Deploy the application 1-baby.yaml, and then change the image baby.jpg to a different one, but keep the same file name baby.jpg.

You can use this image - baby-new.jpg, taken from here - https://pixabay.com/photos/cat-baby-baby-cat-kitten-cat-cute-4201051/

The main goal is to apply a change of a running application and get immediate feedback.

Hints

Finding The Pod/Container - Expand/Collapse

The easiest option would be to use LENS (for this whole challenge). Let's try something else, check the official kubectl cheat sheet to figure out how to get the names of the running Pods.

It's also possible to access minikube's Docker Daemon, which in turn will assist with debugging running containers. This should be used when no other option is available, check this challenge's Solution to see what I mean.

Get In The Container/Pod - Expand/Collapse

First, you need to find the pod's name (see previous hint). Since we have a single container, there's no need to use -c $CONTAINER_NAME. Following that, you can execute

kubectl exec -it "$POD_NAME" -- bash

If you're going to use the above command, you'll get the error "bash": executable file not found. Figure out which command can be used to exec into the running container.

Solution

Expand/Collapse
  1. Find the baby container - instead of grep you can use kubectl get pods --field-selector ..., though I think that using grep is simpler.

    kubectl get po | grep baby
    # baby-79d4f4c875-rb97c                   1/1     Running   0          41m
  2. Exec into baby

    POD_NAME="baby-79d4f4c875-rb97c" # <-- Change this
    kubectl exec -it "$POD_NAME" -- sh

    I used sh because bash is not available in the running container. That is a common practice for Alpine-based images.

  3. Get in the /usr/src/app/images directory and make a backup of baby.jpg

    # In `baby` container
    /usr/src/app $ ls
    # Dockerfile         favicon.ico        index.html         package-lock.json  server.js
    # README.md          images             node_modules       package.json
    
    # Make a backup of current `baby.jpg`
    cd images/
    /usr/src/app/images $ cp baby.jpg baby.jpg.bak
  4. Download the image baby-ch2.jpg and rename it to baby.jpg. The curl application is not available in the container, at least not out of the box.

    It's possible to install it with apt update && apt install -y curl because you're logged in as root. A better alternative is to use wget, which is available out of the box in the running container. Also common for Alpine-based images.

    # -O = output filename 
    IMAGE_URL="https://d33vo9sj4p3nyc.cloudfront.net/kubernetes-localdev/baby-ch2.jpg"
    wget -O baby.jpg "$IMAGE_URL"
  5. Reload the page baby.kubemaster.me; you should see the newly downloaded baby cat image.

    results-ch2.png

Conclusion

Upon completing the challenge - Expand/Collapse

To apply an instant change of a running container, use the command kubectl exec. Instant changes should be done only in development environments since the changes are not permanent. Otherwise, once the Pod (container) restarts, the changes will be gone. Also, applying an immediate change without committing to a git repository makes it impossible to track changes.

The container baby is running as root, this means you can do ANYTHING. That's very bad in production environments, where the blast radius should be minimal. Think about it, if someone hacks into your application, they can download malicious code and access sensitive data with elevated permissions. To read more about "how to run a container as non-root user", check my blog post Docker Tips And Best Practices.

Last but not least, assuming that baby runs with a non-root user, for example, nobody. Assuming you have full access to the Kubernetes Cluster (and Docker Daemon), how would you exec into a container with a root user?

Exec As Root - Expand/Collapse

Unfortunately, there's no out-of-the-box solution for kubectl exec to use a different user. So I'm going to use docker exec --user.

# Use minikube's Docker Daemon
eval `minikube docker-env`

# Get container name
docker ps --filter "name=cats_baby"

# CONTAINER ID   IMAGE          COMMAND                  CREATED             STATUS             PORTS     NAMES
# f72482ace496   da4ce6f08470   "docker-entrypoint.s…"   About an hour ago   Up About an hour             k8s_cats_baby-79d4f4c875-rb97c_default_f8113de0-33be-4ef7-94f2-c3cf6ba98e6b_0

# Mine is: f72482ace496
CONTAINER_NAME="f72482ace496" # <-- Change this

# Check for available users in the running container
docker exec -it "$CONTAINER_NAME" cat /etc/passwd

# root:x:0:0:root:/root:/bin/ash
# ...
# nobody:x:65534:65534:nobody:/:/sbin/nologin

# Exec into the container with `nobody` or `root`
docker exec -it --user root "$CONTAINER_NAME"
# docker exec -it --user nobody "$CONTAINER_NAME"