Yuptime (Project)

Overview

Yuptime is a Kubernetes-native monitoring controller where every monitor is a CRD. There’s no database; state lives on each CRD’s status subresource, and a bounded controller check engine schedules work through persistent workers and a checker sidecar. That makes the whole thing GitOps-native: declare your monitors as YAML in a repo, sync with Flux or Argo CD, or install the public Helm and Timoni OCI artifacts.

Features

All configuration is YAML in Git; works cleanly with Flux, Argo CD, or any GitOps flow
No database: state lives on CRD status subresources
Persistent workers run bounded checks without creating a Kubernetes object for every check
Prometheus metrics for monitoring and alerting
Native Alertmanager integration
Maintenance-window CRDs suppress alerts during planned work

Monitor Types

Network

HTTP: check status code and response body
TCP: port reachability
DNS: resolve and validate records
Ping: ICMP reachability
WebSocket: connect and exchange a frame

Application

Push: heartbeat endpoint for cron jobs and batch processes
Steam: game-server availability
Kubernetes: pod and service health inside the cluster

Architecture

The controller check engine follows a bounded Kubernetes operator pattern:

Reconcilers for Monitor, MonitorSet, and MaintenanceWindow
A fixed worker pool and checker sidecar that schedule and run checks without per-check Jobs
The controller is the ordered status writer; workers hold no database state
A validation layer that rejects malformed CRDs before they hit the reconciler

Configuration

yaml

apiVersion: yuptime.io/v1
kind: Monitor
metadata:
  name: api-health
spec:
  type: http
  enabled: true
  schedule:
    intervalSeconds: 60
    timeoutSeconds: 30
  target:
    http:
      url: https://api.example.com/health

Technology Stack

TypeScript on Bun
Kubernetes client for CRD management
Custom reconciler framework
Prometheus metrics and Alertmanager integration