Compute Inventory
POS-related EC2 instances only. Other Alberaber products (alberaber-erp-tsf, alberaber-infra) are filtered out.
Monthly POS spend (est.)
2 hosts · 1 pool + 1 ops
Estimated — no AWS Billing API; based on on-demand list price.
Current tenants
loading…
Live — from _evrenos.tenant_registry
vs Supabase Cloud Pro
Roughly 70% cheaper than equivalent Cloud projects at our scale.
i-0f09a2a79054c6f38
on-demand
Single shared pool — hosts staging + production tenants on the same EC2. Active tenants (from registry): ucan-balik (production) + kahve-staging-1 (staging). Note: ucan-balik carries the legacy pool_host label 'original-ec2' in the registry; it is physically on this host. Split into a dedicated prod pool when a 2nd paying customer onboards OR memory sustained > 75% for 1 h.
i-04155f702675e42e9
on-demand
Loki + Prometheus + admin dashboard backend for the entire POS fleet. No tenant traffic; serves operators only. Separated from pool-1 so observability stays up when pool-1 is under stress (the moments you most need to see logs).
Per-tenant cost & usagecost = est. share · DB size = real · CPU/mem = cadvisor (null when Prometheus unavailable)
When to split into a dedicated prod pool
We deliberately did not provision prod-pool-1 yet. Triggers that would change that decision:
- 2nd paying customer onboards (blast-radius matters)
- pool-1 memory sustained > 50% for 1 h
- pool-1 swap usage detected during business hours
- Compliance audit demands physical isolation
- You want to do heavy dev experiments without affecting Kahve
Until one of these is true, we stay lean at 2 hosts (~$72.49/mo est.).