Popeye - A Kubernetes Cluster Sanitizer

Popeye is a utility that scans live Kubernetes cluster and reports potential issues with deployed resources and configurations. It sanitizes your cluster based on what's deployed and not what's sitting on disk. By scanning your cluster, it detects misconfigurations and helps you to ensure that best practices are in place, thus preventing future headaches. It aims at reducing the cognitive overload one faces when operating a Kubernetes cluster in the wild. Furthermore, if your cluster employs a metric-server, it reports potential resources over/under allocations and attempts to warn you should your cluster run out of capacity.

Popeye is a readonly tool, it does not alter any of your Kubernetes resources in any way!

---

---

Installation

Popeye is available on Linux, OSX and Windows platforms.

• Binaries for Linux, Windows and Mac are available as tarballs in the release page.

• For OSX/Unit using Homebrew/LinuxBrew

  brew install derailed/popeye/popeye

• Building from source Popeye was built using go 1.12+. In order to build Popeye from source you must:

  1. Clone the repo

  2. Add the following command in your go.mod file

     replace (
       github.com/derailed/popeye => MY_POPEYE_CLONED_GIT_REPO
     )
     

  3. Build and run the executable

     go run main.go

  Quick recipe for the impatient:

  # Clone outside of GOPATH
  git clone https://github.com/derailed/popeye
  cd popeye
  # Build and install
  go install
  # Run
  popeye

PreFlight Checks

• Popeye uses 256 colors terminal mode. On `Nix system make sure TERM is set accordingly.

  export TERM=xterm-256color

Sanitizers

Popeye scans your cluster for best practices and potential issues. Currently, Popeye only looks at nodes, namespaces, pods and services. More will come soon! We are hoping Kubernetes friends will pitch'in to make Popeye even better.

The aim of the sanitizers is to pick up on misconfigurations, i.e. things like port mismatches, dead or unused resources, metrics utilization, probes, container images, RBAC rules, naked resources, etc...

Popeye is not another static analysis tool. It runs and inspect Kubernetes resources on live clusters and sanitize resources as they are in the wild!

Here is a list of some of the available sanitizers:

	Resource	Sanitizers	Aliases
🛀	Node		no
		Conditions ie not ready, out of mem/disk, network, pids, etc
		Pod tolerations referencing node taints
		CPU/MEM utilization metrics, trips if over limits (default 80% CPU/MEM)
🛀	Namespace		ns
		Inactive
		Dead namespaces
🛀	Pod		po
		Pod status
		Containers statuses
		ServiceAccount presence
		CPU/MEM on containers over a set CPU/MEM limit (default 80% CPU/MEM)
		Container image with no tags
		Container image using latest tag
		Resources request/limits presence
		Probes liveness/readiness presence
		Named ports and their references
🛀	Service		svc
		Endpoints presence
		Matching pods labels
		Named ports and their references
🛀	ServiceAccount		sa
		Unused, detects potentially unused SAs
🛀	Secrets		sec
		Unused, detects potentially unused secrets or associated keys
🛀	ConfigMap		cm
		Unused, detects potentially unused cm or associated keys
🛀	Deployment		dp, deploy
		Unused, pod template validation, resource utilization
🛀	StatefulSet		sts
		Unsed, pod template validation, resource utilization
🛀	DaemonSet		ds
		Unsed, pod template validation, resource utilization
🛀	PersistentVolume		pv
		Unused, check volume bound or volume error
🛀	PersistentVolumeClaim		pvc
		Unused, check bounded or volume mount error
🛀	HorizontalPodAutoscaler		hpa
		Unused, Utilization, Max burst checks
🛀	PodDisruptionBudget
		Unused, Check minAvailable configuration	pdb
🛀	ClusterRole
		Unused	cr
🛀	ClusterRoleBinding
		Unused	crb
🛀	Role
		Unused	ro
🛀	RoleBinding
		Unused	rb
🛀	Ingress
		Valid	ing
🛀	NetworkPolicy
		Valid	np
🛀	PodSecurityPolicy
		Valid	psp

Save the report

To save the Popeye report to a file pass the --save flag to the command. By default it will create a temp directory and will store the report there, the path of the temp directory will be printed out on STDOUT. If you have the need to specify the output directory for the report, you can use the environment variable POPEYE_REPORT_DIR.

Example to save report in working directory:

  $ POPEYE_REPORT_DIR=$(pwd) popeye --save

Save the report to S3

You can also save the generated report to an AWS S3 bucket with providing the flag --s3-bucket. As parameter you need to provide the name of the S3 bucket where you want to store the report. To save the report in a bucket subdirectory provide the bucket parameter as bucket/path/to/report.

Underlying the AWS Go lib is used which is handling the credential loading. For more information check out the official documentation.

Example to save report to S3:

  $ popeye --s3-bucket=NAME-OF-YOUR-S3-BUCKET/OPTIONAL/SUBDIRECTORY --out=json

Run public docker image locally

You don't have to build and/or install the binary to run popeye: you can just run it directly from the official docker repo on quay.io. The default command when you run the docker container is popeye, so you just need to pass whatever cli args are normally passed to popeye. To access your clusters, map your local kube config directory into the container with -v :

  $ docker run --rm -it -v $HOME/.kube:/root/.kube \
      quay.io/derailed/popeye:latest --context foo -n bar

Running the above docker command with --rm means that the container gets deleted when popeye exits. When you use --save, it will write it to /tmp in the container and then delete the container when popeye exits, which means you lose the output. To get around this, map /tmp to the container's /tmp:

  $ docker run --rm -it -v $HOME/.kube:/root/.kube -v /tmp:/tmp \
      quay.ioderailed/popeye:latest --context foo -n bar --save
  /tmp/popeye/sanitizer_foo_1583042166995001900.txt

  # Docker has exited, and the container has been deleted, but the file
  # is in your /tmp subdir because you mapped it into the container
  $ cat /tmp/popeye/sanitizer_foo_1583042166995001900.txt
    <snip>

The Command Line

You can use Popeye standalone or using a spinach yaml config to tune the sanitizer. Details about the Popeye configuration file are below.

# Dump version info
popeye version
# Popeye a cluster using your current kubeconfig environment.
popeye
# Popeye uses a spinach config file of course! aka spinachyaml!
popeye -f spinach.yml
# Popeye a cluster using a kubeconfig context.
popeye --context olive
# Stuck?
popeye help

Output Formats

Popeye can generate sanitizer reports in a variety of formats. You can use the -o cli option and pick your poison from there.

Format	Description	Default	Credits
standard	The full monty output iconized and colorized	yes
jurassic	No icons or color like it's 1989
yaml	As YAML
html	As HTML
json	As JSON
junit	For the Java melancholic
prometheus	Dumps report a prometheus scrappable metrics		dardanel
score	Returns a single cluster sanitizer score value (0-100)		kabute

The SpinachYAML Configuration

A spinach.yml configuration file can be specified via the -f option to further configure the sanitizers. This file may specify the container utilization threshold and specific sanitizer configurations as well as resources that will be excluded from the sanitization.

NOTE: This file will change as Popeye matures!

A resource is identified by a resource kind and a fully qualified resource name, i.e. namespace/resource_name.

For example, the FQN of a pod named fred-1234 in the namespace blee will be blee/fred-1234. This provides for differentiating fred/p1 and blee/p1. For cluster wide resources, the FQN is equivalent to the name. Exclude rules can have either a straight string match or a regular expression. In the latter case the regular expression must be indicated using the rx: prefix.

NOTE! Please be careful with your regex as more resources than expected may get excluded from the report with a loose regex rule. When your cluster resources change, this could lead to a sub-optimal sanitization. Once in a while it might be a good idea to run Popeye „configless“ to make sure you will recognize any new issues that may have arisen in your clusters…

Here is an example spinach file as it stands in this release. There is a fuller eks and aks based spinach file in this repo under spinach. (BTW: for new comers into the project, might be a great way to contribute by adding cluster specific spinach file PRs...)

# A Popeye sample configuration file
popeye:
  # Checks resources against reported metrics usage.
  # If over/under these thresholds a sanitization warning will be issued.
  # Your cluster must run a metrics-server for these to take place!
  allocations:
    cpu:
      underPercUtilization: 200 # Checks if cpu is under allocated by more than 200% at current load.
      overPercUtilization: 50   # Checks if cpu is over allocated by more than 50% at current load.
    memory:
      underPercUtilization: 200 # Checks if mem is under allocated by more than 200% at current load.
      overPercUtilization: 50   # Checks if mem is over allocated by more than 50% usage at current load.

  # Excludes excludes certain resources from Popeye scans
  excludes:
    # ConfigMap sanitizer exclusions...
    configmap:
      # Excludes key must match the singular form of the resource.
      # For instance this rule will exclude all configmaps named fred.v2.3 and fred.v2.4
      - name: rx:fred.+\.v\d+
    # Namespace sanitizer exclusions...
    namespace:
      # Exclude all fred* namespaces if the namespaces are not found (404), other error codes will be reported!
      - name: rx:kube
        codes:
          - 404
      # Exclude all istio* namespaces from being scanned.
      - name: rx:istio

  # Configure node resources.
  node:
    # Limits set a cpu/mem threshold in % ie if cpu|mem > limit a lint warning is triggered.
    limits:
      # CPU checks if current CPU utilization on a node is greater than 90%.
      cpu:    90
      # Memory checks if current Memory utilization on a node is greater than 80%.
      memory: 80

  # Configure pod resources
  pod:
    # Restarts check the restarts count and triggers a lint warning if above threshold.
    restarts:
      3
    # Check container resource utilization in percent.
    # Issues a lint warning if about these threshold.
    limits:
      cpu:    80
      memory: 75

Popeye In Your Clusters!

Alternatively, Popeye is containerized and can be run directly in your Kubernetes clusters as a one-off or CronJob.

Here is a sample setup, please modify per your needs/wants. The manifests for this are in the k8s directory in this repo.

kubectl apply -f k8s/popeye/ns.yml && kubectl apply -f k8s/popeye

---
apiVersion: batch/v1
kind: CronJob
metadata:
  name:      popeye
  namespace: popeye
spec:
  schedule: "* */1 * * *" # Fire off Popeye once an hour
  concurrencyPolicy: Forbid
  jobTemplate:
    spec:
      template:
        spec:
          serviceAccountName: popeye
          restartPolicy: Never
          containers:
            - name: popeye
              image: quay.io/derailed/popeye:v0.8.6
              imagePullPolicy: IfNotPresent
              args:
                - -o
                - yaml
              resources:
                limits:
                  cpu:    500m
                  memory: 100Mi

Popeye got your RBAC!

In order for Popeye to do his work, the signed-in user must have enough RBAC oomph to get/list the resources mentioned above.

Sample Popeye RBAC Rules (please note that those are subject to change.)

---
# Popeye ServiceAccount.
apiVersion: v1
kind:       ServiceAccount
metadata:
  name:      popeye
  namespace: popeye

---
# Popeye needs get/list access on the following Kubernetes resources.
apiVersion: rbac.authorization.k8s.io/v1
kind:       ClusterRole
metadata:
  name: popeye
rules:
- apiGroups: [""]
  resources:
   - configmaps
   - deployments
   - endpoints
   - horizontalpodautoscalers
   - namespaces
   - nodes
   - persistentvolumes
   - persistentvolumeclaims
   - pods
   - secrets
   - serviceaccounts
   - services
   - statefulsets
  verbs:     ["get", "list"]
- apiGroups: ["rbac.authorization.k8s.io"]
  resources:
  - clusterroles
  - clusterrolebindings
  - roles
  - rolebindings
  verbs:     ["get", "list"]
- apiGroups: ["metrics.k8s.io"]
  resources:
  - pods
  - nodes
  verbs:     ["get", "list"]

---
# Binds Popeye to this ClusterRole.
apiVersion: rbac.authorization.k8s.io/v1
kind:       ClusterRoleBinding
metadata:
  name: popeye
subjects:
- kind:     ServiceAccount
  name:     popeye
  namespace: popeye
roleRef:
  kind:     ClusterRole
  name:     popeye
  apiGroup: rbac.authorization.k8s.io

Screenshots

Cluster D Score

Cluster A Score

Report Morphology

The sanitizer report outputs each resource group scanned and their potential issues. The report is color/emoji coded in term of Sanitizer severity levels:

Level	Icon	Jurassic	Color	Description
Ok	✅	OK	Green	Happy!
Info	🔊	I	BlueGreen	FYI
Warn	😱	W	Yellow	Potential Issue
Error	💥	E	Red	Action required

The heading section for each scanned Kubernetes resource provides a summary count for each of the categories above.

The Summary section provides a Popeye Score based on the sanitization pass on the given cluster.

Known Issues

This initial drop is brittle. Popeye will most likely blow up when…

• You're running older versions of Kubernetes. Popeye works best with Kubernetes 1.13+.
• You don't have enough RBAC oomph to manage your cluster (see RBAC section)

Disclaimer

This is work in progress! If there is enough interest in the Kubernetes community, we will enhance per your recommendations/contributions. Also if you dig this effort, please let us know that too!

ATTA Girls/Boys!

Popeye sits on top of many of open source projects and libraries. Our sincere appreciations to all the OSS contributors that work nights and weekends to make this project a reality!

Contact Info

1. Email: [email protected]
2. Twitter: @kitesurfer

---

 © 2020 Imhotep Software LLC. All materials licensed under Apache v2.0