Ephemeral compute with persistent test data
Why preview environments should destroy servers aggressively while protecting the expensive part of test setup: known data.
“Ephemeral environment” is often interpreted as “delete everything.” That is safe for compute, but wasteful when a realistic database takes hours to import, sanitize, or seed.
Two resources, two decisions
Compute and data have different reasons to exist:
- Compute is useful only while someone is testing, demonstrating, or debugging.
- Seeded data remains useful across many of those sessions.
Putting both into one destroy operation forces a bad choice: pay for an idle server or repeat an expensive seed.
A protected data stack
Chalupa keeps persistent block storage in a dedicated <stack>-data Pulumi stack. The disposable compute stack consumes its exact exported volume identifier. It never guesses by resource name, and an ordinary chalupa down never targets the data stack.
Permanent data deletion is deliberately absent from the CLI surface. It remains an explicit task data-nuke repository operation with its own typed confirmation; chalupa down never targets the data stack.
The cost model becomes legible
When compute is down, its hourly cost falls to zero. The volume continues to incur a small, predictable monthly charge until a human decides the seed is no longer valuable. The control plane labels both as estimates; the provider invoice remains authoritative.
Read the full data and seeding guide or see pricing for the hosted control plane.