Skip to content

GCP Servers

Core Infrastructure

All core infrastructure runs on GCP Compute Engine VMs in europe-west8, managed via Terraform + Terragrunt. Each environment has a dedicated Backend VM and Compute VM.

OS image: Ubuntu 22.04 LTS on all VMs. Boot disks are set to auto_delete = false.

Production

GCP Project: sartiq-prod

production-backend
Machine Type n2-standard-4 (4 vCPU / 16 GB)
Zone europe-west8-b
Internal IP 10.0.0.2
Boot Disk 20 GB SSD (pd-ssd)
Data Disk 100 GB SSD (pd-ssd)
Network Tags ssh, http, backend
Service Account production-backend@sartiq-prod
Deploy Path /home/ubuntu/services/shootify-deploy

Services: Caddy, Backend API, Scheduler, Celery Worker, PostgreSQL, Redis, DB Backup

URLs: - https://api.sartiq.com — Backend API - https://compute-api.sartiq.com — Compute API (proxied via Caddy)

production-compute
Machine Type n2-standard-8 (8 vCPU / 32 GB)
Zone europe-west8-b
Internal IP 10.0.0.3
Boot Disk 20 GB SSD (pd-ssd)
Data Disk 200 GB SSD (pd-ssd)
Network Tags ssh, compute
Service Account production-compute@sartiq-prod
Deploy Path /home/ubuntu/services/shootify-deploy/compute

Services: Compute API, FAL Worker, LLM Worker, Local CPU Worker, Workflow Worker, Inference Worker, Beat, Event Listener, PostgreSQL, Redis, DB Backup

Staging

GCP Project: sartiq-staging

staging-backend
Machine Type e2-medium (2 vCPU / 4 GB)
Zone europe-west8-b
Internal IP 10.0.1.3
Boot Disk 20 GB SSD (pd-ssd)
Data Disk 50 GB SSD (pd-ssd)
Network Tags ssh, http, backend
Service Account staging-backend@sartiq-staging
Deploy Path /home/ubuntu/services/shootify-deploy

Services: Caddy, Backend API, Scheduler, Celery Worker, Webapp, PostgreSQL, Redis, DB Backup

URLs: - https://staging-api.sartiq.com — Backend API - https://compute-api-staging.sartiq.com — Compute API (proxied via Caddy)

staging-compute
Machine Type n2-standard-4 (4 vCPU / 16 GB)
Zone europe-west8-b
Internal IP 10.0.1.4
Boot Disk 20 GB SSD (pd-ssd)
Data Disk 100 GB SSD (pd-ssd)
Network Tags ssh, compute
Service Account staging-compute@sartiq-staging
Deploy Path /home/ubuntu/services/shootify-deploy/compute

Services: Compute API, FAL Worker, LLM Worker, Local CPU Worker, Workflow Worker, Inference Worker, Beat, Event Listener, PostgreSQL, Redis, DB Backup

Development

GCP Project: sartiq-staging (shares project with staging)

dev-backend
Machine Type e2-medium (2 vCPU / 4 GB)
Zone europe-west8-b
Internal IP 10.0.2.3
Boot Disk 20 GB SSD (pd-ssd)
Data Disk 30 GB SSD (pd-ssd)
Network Tags ssh, http, backend
Service Account dev-backend@sartiq-staging
Deploy Path /home/ubuntu/services/shootify-deploy

Services: Caddy, Backend API, Scheduler, Celery Worker, Webapp, PostgreSQL, Redis, DB Backup

URLs: - https://dev-api.sartiq.com — Backend API - https://compute-api-dev.sartiq.com — Compute API (proxied via Caddy)

dev-compute
Machine Type e2-standard-2 (2 vCPU / 8 GB)
Zone europe-west8-c
Internal IP 10.0.2.2
Boot Disk 20 GB SSD (pd-ssd)
Data Disk 50 GB SSD (pd-ssd)
Network Tags ssh, compute
Service Account dev-compute@sartiq-staging
Deploy Path /home/ubuntu/services/shootify-deploy/compute

Services: Compute API, FAL Worker, LLM Worker, Local CPU Worker, Workflow Worker, Inference Worker, Beat, Event Listener, PostgreSQL, Redis, DB Backup

Note: Dev compute is in zone europe-west8-c, separate from the backend in zone b.

Data Disk Mounting

All VMs mount a persistent SSD at /data (ext4, nofail). Data disks have prevent_destroy = true in Terraform.

/data/
├── postgres/          # PostgreSQL data directory
├── redis/             # Redis snapshots
├── caddy/             # Caddy config and certificates (backend only)
│   ├── data/
│   └── config/
├── config/            # Deployment config symlinks (.env, docker-compose.yml, Caddyfile)
└── models/onnx/       # ML models (compute only)

Docker is installed via the VM startup script and the default user (ubuntu) is added to the docker group.

Supporting Services

sartiq-dev-01
Machine Type e2-standard-2
Zone europe-west8-b
External IP 34.154.18.83
Role Hosts dev/experimental web applications

URLs: https://*.dev.sartiq.com

GPU / Experimental Servers

These servers are managed manually (not via Terraform) and used ad-hoc by the team for ML experiments and training.

Name Address GPUs URLs
dev-a100-80 34.28.243.244 1x A100 80GB https://ai-toolkit.sartiq.com
labs-01-2x-rtxpro6000-96g 34.28.243.244 2x RTXPRO6000 96GB https://ai-toolkit-1.sartiq.com