Skip to content

infrastructure-index.md

Latest commit

 

History

History
63 lines (48 loc) · 2.8 KB

infrastructure-index.md

File metadata and controls

63 lines (48 loc) · 2.8 KB

Infrastructure Index

This index documents Ampersand's infrastructure, deployment systems, and operations.

Major Components

Google Cloud Platform

  • GCP Infrastructure - GKE, Cloud SQL, Pub/Sub, Cloud Functions, KMS, storage, networking --> infrastructure/gcp-infrastructure.md

Kubernetes & Service Mesh

  • GKE cluster per environment gradient (prod/staging/dev/preview)
  • Kong API Gateway (internet-facing + outpost-facing)
  • Service collapsing in dev (single pod handles api/read/write/proxy)
  • Pod Disruption Budgets for HA in production
  • See infrastructure/gcp-infrastructure.md for details

CI/CD & Deployment

  • Deployment Pipeline - Cloud Build, ArgoCD GitOps, promotion flow, preview environments --> infrastructure/deployment-pipeline.md
  • Testing onion: unit -> integration -> preview -> dev -> staging -> prod
  • Sed-based templating (not Helm/Kustomize)
  • Promotion: dev -> staging -> prod (no skipping)

Databases

  • Database Architecture - PostgreSQL via GORM, Atlas migrations, repository pattern, entity hierarchy --> infrastructure/database-architecture.md
  • PostgreSQL 15, 107+ migrations, 64-core in prod
  • Five-package structure: types/models/gormdb/memory/mock

Message Queues & Pub/Sub

8 primary Pub/Sub topics for inter-service communication:

Topic Publisher Subscriber Purpose
read-finished Temporal Messenger Completed read operations
poll-bulk-job Temporal Messenger Bulk write job status
subscribe-webhook-received Cloud Functions Messenger Incoming provider webhooks
subscribe-webhook-processed Messenger Messenger Processed events for delivery
subscribe-async-installation API Temporal Async installation events
usage-emitted Various Scribe Billing/usage events
operation-change Various Scribe Operation status updates
notification-emitted Various Messenger Notification events

See infrastructure/gcp-infrastructure.md for full Pub/Sub details.

Webhook Delivery

  • Svix (managed) - Webhook destinations
  • Outpost (self-hosted, Hookdeck) - Kinesis destinations, at-least-once delivery
  • Future: consolidation to Outpost for all delivery

Environment Gradient

Environment Services DB Cores Auto-scaling Alerting
Production Full separation (10 deployments) 64 Yes PagerDuty, Slack
Staging Some separation Mid-tier Yes Conditional
Dev Collapsed (1 API pod) Minimal No None
Preview Collapsed (like dev) Shared dev DB No None

Cross-Cutting Concerns

  • Services architecture --> See services-index.md
  • Security --> See security-index.md
  • Observability --> See observability-index.md