Skip to content

README.md

Latest commit

 

History

History
86 lines (62 loc) · 4.17 KB

README.md

File metadata and controls

86 lines (62 loc) · 4.17 KB

worker — InstaNode background-job processor

Part of the InstaNode platform. This repository runs the asynchronous side of the platform — the jobs that don't fit in a request handler because they take too long, need to retry on failure, or run on a schedule.

What this does

  • Reaps anonymous + free-tier resources at their TTL, with warning emails at 6h / 2h / 1h
  • Reconciles tier-grade changes after a Razorpay subscription event (entitlement_reconciler)
  • Scans per-resource storage usage every minute and emits 80% upgrade nudges (quota_wall_nudge)
  • Reaps orphan deploy/stack namespaces left behind by failed builds (orphan_sweep_reconciler, 6 passes)
  • Runs the eager-retry consumer for pending_propagations with bounded backoff + dead-letter (propagation_runner)
  • Backs up customer postgres + mongodb databases on a schedule (customer_backup_runner); restores on demand (customer_restore_runner)
  • Forwards every audit_log row that has an associated email template to Brevo + records the delivery classification on the forwarder_sent ledger
  • Repairs Razorpay subscription/checkout state (billing_reconciler, checkout_reconcile) and handles payment-grace expiry (payment_grace_terminator)
  • Synthesises uptime probes against customer Postgres/Redis/Mongo (real_prober, uptime_prober)

Job-runner is River (Postgres-native queue), scheduled by River cron + manual triggers from the api repo.

Architecture

api/  ───emits──>  audit_log + pending_propagations
                          │
                          ▼
worker/  ───consumes────  River queue + scheduled jobs
                          │
                          ├──> Brevo (email)
                          ├──> Razorpay (subscription state)
                          ├──> provisioner gRPC (k8s namespace teardown)
                          └──> DigitalOcean Spaces (backup storage)

worker shares Go types and helpers with api and provisioner through the common repo (queue providers, storage providers, plans registry, readiness checks, logging context).

Quick start

git clone https://github.com/InstaNode-dev/worker
cd worker
go build ./...
go vet ./...
go test ./... -short -p 1

Local development requires postgres (platform DB + customer DB), redis, and access to the api repo's customer_backup_scheduler table schema. See infra/docker-compose.yml for a one-command local stack.

Configuration

Driven entirely by environment variables — internal/config/config.go is the source of truth. Key ones:

  • DATABASE_URL — platform DB connection string (sslmode=require for managed PG)
  • CUSTOMER_DATABASE_URL — customer DB connection (separate cluster)
  • REDIS_URL — platform Redis
  • PROVISIONER_ADDR — gRPC endpoint of the provisioner service
  • BREVO_API_KEY — transactional email
  • RAZORPAY_KEY_ID / RAZORPAY_KEY_SECRET — billing
  • OBJECT_STORE_BACKEND + OBJECT_STORE_* — DO Spaces / R2 / S3 / MinIO selection (see common/storageprovider/)
  • NEW_RELIC_LICENSE_KEY — observability

See .env.example for the full list and safe development defaults.

Running locally

DATABASE_URL=postgres://localhost/instant_platform?sslmode=disable \
  REDIS_URL=redis://localhost:6379 \
  go run .

The worker starts a River pool, registers ~50 job kinds, and serves a deep /readyz on :8091 (covers platform_db, redis-provision, brevo reachability, River health) and a shallow /healthz.

Observability

  • Prometheus metrics at /metrics on :8091 — see infra/observability/METRICS-CATALOG.md for the full registry
  • structured JSON logs via slog with request_id + commit_id propagation (common/logctx)
  • OpenTelemetry traces exported to New Relic (when NEW_RELIC_LICENSE_KEY is set)

Contributing

See CONTRIBUTING.md. License is MIT — see LICENSE.

For platform-wide issues, file at the api repo. For worker-specific bugs, file here.

Security

See SECURITY.md for the responsible-disclosure policy.