README
¶
k8s-metacollector
⚠️ The repository is still a work in progress ⚠️
The "K8s Meta Collector" is a self-contained module that can be deployed within a Kubernetes cluster to perform the task of gathering metadata from various Kubernetes resources and subsequently transmitting this collected metadata to designated subscribers.
Description
Khulnasoft enriches events coming from syscall event source with metadata
coming from other sources, for example Kubernetes API server. Historically, each instance of Khulnasoft running in a
Kubernetes cluster would connect to the Kubernetes API server in order to fetch the metadata for a subset of
Kubernetes resources. This approach works
well in small Kubernetes cluster but does not scale in large environments. The following issue describes the
problems that were affecting the old Kubernetes client: https://github.com/khulnasoft/libs/issues/987.
The aim of k8s-meta-collector is to propose a novel approach to k8s metadata enrichment in Khulnasoft by moving
the fetching logic of the metadata to a centralized component. The Khulnasoft instances would connect to this component
and receive the metadata without the need to connect to the Kubernetes API server.
The following image shows the deployment of k8s-meta-collector and Khulnasoft in a kubernetes cluster.
Having a centralized component that connects to the API server and pushes metadata to the Khulnasoft instances reduces the load on the Kubernetes API server. Keep in mind that Khulnasoft is deployed as a DaemonSet, one Khulnasoft instance on each node. It also reduces the number of events sent to the Khulnasoft instances by filtering the metadata by the node. A given Khulnasoft instance running in a given node will receive metadata only for the resources that are related to that node:
- pods running on the node;
- namespaces that contain a pod running on the node;
- deployment, replicaset, replicationcontrollers associated with a pod running on the node;
- services serving a pod running on the node.
The filtering done by k8s-meta-collector reduces significantly the number of events sent to the Khulnasoft instances.
The metadata received by the subscribers is ready to be used without the need for further processing on the
subscribers side.
Functional Guarantees:
The k8s-meta-collector assures that:
- subscribers (Khulnasoft instances) at subscribe time will receive all the metadata for the resources related to the subscriber(node for which the subscriber wants to receive the metadata);
- a message of type
Createis sent to the subscribers when a new resource is discovered; for it; - a message of type
Updateis sent to the subscriber when an already sent resource has some fields modified; - a message of type
Deleteis sent to the subscriber when an already sent resource is not anymore relevant for the subscriber; - only metadata for resources related to a subscriber are sent;
Getting Started
You’ll need a Kubernetes cluster to run against. You can use KIND to get a local cluster for testing, or run against a remote cluster.
Running on the cluster
It's as easy as running:
kubectl apply -f manifests/meta-collector.yaml
If you want to scrape the metrics exposed by k8s-meta-collector using prometheus then deploy the provided
ServiceMonitor. Make sure to add the appropriate label to the manifest file in order to be discovered and scraped by
your prometheus instance.
kubectl apply -f manifests/monitor.yaml
There is also a default grafana dashboard ready to be used under grafana folder.
License
This project is licensed to you under the Apache 2.0 license.
Documentation
Source Files
Directories
¶
| Path | Synopsis |
|---|---|
|
Package broker implements the broker logic.
|
Package broker implements the broker logic. |
|
cmd
|
|
|
collector
Package collector implements the command line for the metacollector.
|
Package collector implements the command line for the metacollector. |
|
collector/run
Package run implement the run command.
|
Package run implement the run command. |
|
collector/version
Package version implements the logic for the version command.
|
Package version implements the logic for the version command. |
|
Package collectors implements the collectors' logic.
|
Package collectors implements the collectors' logic. |
|
Package metadata defines the grpc metadata service and the server code.
|
Package metadata defines the grpc metadata service and the server code. |
|
pkg
|
|
|
consts
Package consts provides constants used across the project.
|
Package consts provides constants used across the project. |
|
events
Package events provides the events generated by the collectors.
|
Package events provides the events generated by the collectors. |
|
fields
Package fields contains the types for the fields extracted by k8s resources.
|
Package fields contains the types for the fields extracted by k8s resources. |
|
resource
Package resource provides types for the k8s resources.
|
Package resource provides types for the k8s resources. |
|
subscriber
Package subscriber provides types and data structures for subscribers.
|
Package subscriber provides types and data structures for subscribers. |
|
version
Package version hold the logic for the version.
|
Package version hold the logic for the version. |