Connect Akto eBPF with Kubernetes

Introduction

Connecting with Akto's eBPF traffic collector is recommended for mTLS systems.

For a better understanding, here's an architecture diagram of the setup.

ebpf Deployment

Adding Akto traffic collector

1. Setup Akto data processor using the guide here

2. Apply the Daemonset configuration given below using kubectl apply -f auto-daemon set-config.yaml -n <NAMESPACE>. You will find AKTO_NLB_IP after setting up Akto data processor, as mentioned above.

---
apiVersion: v1
kind: ServiceAccount
metadata:
  name: pod-watcher
  namespace: {NAMESPACE} 

---
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRole
metadata:
  name: pod-watcher-role
rules:
- apiGroups: [""]
  resources: ["pods"]
  verbs: ["get", "watch", "list"]

---
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRoleBinding
metadata:
  name: pod-watcher-role-binding
subjects:
- kind: ServiceAccount
  name: pod-watcher
  namespace: {NAMESPACE}
roleRef:
  kind: ClusterRole
  name: pod-watcher-role
  apiGroup: rbac.authorization.k8s.io
---
apiVersion: apps/v1
kind: DaemonSet
metadata:
  name: akto-k8s
  namespace: {NAMESPACE}
  labels:
    app: akto-collector
spec:
  selector:
    matchLabels:
      app: akto-collector
  template:
    metadata:
      labels:
        app: akto-collector
    spec:
      hostNetwork: true
      serviceAccountName: pod-watcher
      dnsPolicy: ClusterFirstWithHostNet
      hostPID: true
      containers:
      - name: mirror-api-logging
        image: aktosecurity/mirror-api-logging:k8s_ebpf
        resources:
          limits:
            cpu: 500m
            memory: 1Gi
          requests:
            cpu: 50m
            memory: 50Mi
        env: 
          - name: AKTO_TRAFFIC_BATCH_TIME_SECS
            value: "10"
          - name: AKTO_TRAFFIC_BATCH_SIZE
            value: "100"
          - name: AKTO_KAFKA_BROKER_MAL
            value: "<AKTO_NLB_IP>:9092"
          - name: AKTO_K8_METADATA_CAPTURE
            value: "true"
          - name: NODE_NAME
            valueFrom:
              fieldRef:
                fieldPath: spec.nodeName
        securityContext:
          capabilities:
            add:
            - SYS_PTRACE
            - SYS_ADMIN
          privileged: true
        volumeMounts:
          # needed to load kernel headers
          - name: lib-modules
            mountPath: /lib/modules
            readOnly: true
          # needed to trace kernel events
          - name: sys-kernel
            mountPath: /sys/kernel
            readOnly: true
          - name: host
            mountPath: /host
            readOnly: true
      volumes:
        - name: sys-kernel
          hostPath:
            path: /sys/kernel
        - name: lib-modules
          hostPath:
            path: /lib/modules
        - name: host
          hostPath:
            path : /

1. You can add and configure the env variables below to control the daemonset. Here's a diagram of how the module processes traffic:

eBPF Traffic Processing

# This helps in filtering traffic sent to akto, based on certain headers. Here is an example for sending traffic only for 'bookinfo' namespace in an istio setup.
- name: AKTO_MODULE_DISCOVERY_CONFIG
  value: '[{"key":{"eq":"x-forwarded-client-cert","ifAbsent":"reject"},"value":{"regex":".*bookinfo.*"}}]'
# Time limit ( in seconds ) after which, a traffic stream is processed and marked inactive. The same stream, is not processed again.
- name: TRAFFIC_INACTIVITY_THRESHOLD
  value: "30"
# Max traffic connections kept in memory 
- name: TRAFFIC_MAX_ACTIVE_CONN
  value: "4096"
# Make this flag true to disable egress traffic. It is recommended to keep this false.
- name: TRAFFIC_DISABLE_EGRESS
  value: "false"
# Max mem usage after which the pod restarts ( in MB )
- name: AKTO_MEM_THRESH_RESTART
  value: "800"
# Max limit of traffic buffer kept in memory ( in MB )
- name: TRAFFIC_BUFFER_THRESHOLD
  value: "600"
# Ignore traffic coming from unresolved IPs, i.e. requests with host header of the format <a.b.c.d>
- name: AKTO_IGNORE_IP_TRAFFIC
  value: "false"
# Ignore traffic coming from AWS cloud metadata IP
- name: AKTO_IGNORE_CLOUD_METADATA_CALLS
  value: "false"
# The interval poll ( in seconds ) in which data is sent to Akto data processor.
- name: KAFKA_POLL_INTERVAL
  value: "0.5"
# If you only want to trace traffic for which SSL termination happens at proxy/service.
- name: CAPTURE_ALL
  value: "false"

1. You can check your API inventory on Akto dashboard to see endpoints being discovered.

Kubernetes Pod Labels Tagging

Akto traffic collector can be configured to capture Kubernetes Pod labels with each API request. It operates within the Akto eBPF DaemonSet and leverages the Kubernetes Informer to maintain a local in-memory cache of pods running on each node. This cache is used to identify the labels of pod.

Env Variables

# This will start capturing the pod labels from all namespaces except
# kube-system, kube-public and kube-node-lease
- name: AKTO_K8_METADATA_CAPTURE
  value: "true"
# Use this if you want to capture pod labels from a specific namespace only.
- name: AKTO_K8_METADATA_CAPTURE_NAMESPACE
  value: {NAMESPACE}

Frequently Asked Questions (FAQs)

The traffic will contain a lot of sensitive data - does it leave my VPC?

Data remains strictly within your VPC. Akto doesn't take data out of your VPC at all.

Does adding DaemonSet have any impact on performance or latency?

Zero impact on latency. The DaemonSet doesn't sit like a proxy. It works on eBPF technology, which works on traces function calls at kernel level. It is very lightweight. We have benchmarked it against traffic as high as 20M API requests/min. It consumes very low resources (CPU & RAM).

How can I control logs generated by akto's eBPF traffic collector ?

You can utilize the AKTO_LOG_LEVEL environment variable, which accepts DBEUG, INFO, WARN, ERROR, OFF as log levels. Default level is set to WARN.

I don't see my error on this list here.

Please send us all details at [email protected] or reach out via Intercom on your Akto dashboard. We will definitely help you out.

Get Support for your Akto setup

There are multiple ways to request support from Akto. We are 24X7 available on the following:

1. In-app intercom support. Message us with your query on intercom in Akto dashboard and someone will reply.

2. Join our discord channel for community support.

3. Contact [email protected] for email support.

4. Contact us here.