Munchbox
A hybrid homelab platform built on HashiCorp Nomad, Consul, and Vault. The cluster spans bare-metal nodes, Proxmox VMs, and Oracle Cloud free-tier instances connected over WireGuard across amd64 and arm64 architectures. Managed with Terraform and Ansible, monitored with Prometheus, Grafana, Loki, and Tempo.
View on GitHubBuilt for the munchbox
Combines multiple S3-compatible storage backends into a single unified endpoint with per-backend quota enforcement.
- Supports any S3-compatible backend (OCI, Backblaze B2, IDrive e2, MinIO, AWS, etc.)
- Per-backend byte quotas and monthly API/ingress/egress quota enforcement
- Configurable placement strategies (pack, spread)
- Virtual buckets with isolated namespaces and independent credentials
- Optional cross-backend replication/failover and envelope encryption
An S3-compatible HTTP gateway that uses Gmail and Google Drive as the storage backend.
- Object data in Google Drive, metadata as Gmail emails with JSON pointers
- Local SQLite or PostgreSQL index for zero-API-call metadata operations
- Buckets map to Gmail labels, works with AWS CLI and any S3 SDK
- Multipart upload support for large files
- SigV4 authentication, Prometheus metrics, and OpenTelemetry tracing
Temporal workflow workers that automate daily backups, Trivy vulnerability scanning, and stale allocation cleanup.
- Automated PostgreSQL/Consul/Nomad backups to S3
- Trivy vulnerability scanning of all running container images
- Stale allocation cleanup on Nomad client nodes via SSH
- Transient vs. permanent error classification with configurable retry
- OpenTelemetry tracing and Prometheus metrics
Polls the Cloudflare GraphQL Analytics API for firewall events and HTTP traffic, then ships them into Loki, Prometheus, and Tempo.
- WAF/firewall event collection pushed to Loki as structured JSON
- Aggregated HTTP traffic stats exposed as Prometheus gauges
- Automatic log-trace correlation via injected trace_id/span_id
- Cloudflare rate limit handling with Retry-After support
- OpenTelemetry tracing per poll cycle
Detects when Oracle Cloud reclaims free-tier instances and automatically stop/start cycles them back to life via the OCI API.
- Consul session heartbeats with configurable TTL
- Automatic OCI stop/start cycle on missed heartbeats
- Duplicate restart prevention via in-flight tracking
- Configurable max consecutive restart attempts per node
- Dry-run mode for testing