Environments: per-environment deploys
Ship the same project to several places: production, a staging copy, a throwaway dev slice. An environment block declares another slice from the same file, and variable blocks say what differs between them. drift plan staging shows what the staging slice would change, and drift apply staging makes it so, creating the slice when it does not exist.
Each environment is a slice of its own, with its own bill.
prod and staging environments is two slices, links and links-staging, each shaped and priced on its own. They are separate in every sense: their own data, their own URL, their own bill. drift plan staging quotes the staging slice's price before drift apply staging agrees to it.An environment changes only what a variable carries, plus its slice. There is no per-environment override of whole sections and no merging: the file is written once, and a variable stands wherever a value should differ.
slice "links" {}
variable "memory" {
default = "64MiB"
}
variable "store" {
default = "64MiB"
}
variable "rate" {
default = "600/min"
}
variable "seed" {
default = "backbone/seed.jsonl"
}
environment "prod" {}
environment "staging" {
vars = {
memory = "32MiB"
store = "16MiB"
rate = "60/min"
seed = "backbone/demo.jsonl"
}
}
environment "dev" {
slice = "links-sandbox"
vars = { memory = "16MiB" }
}
atomic {
backend "default" {
function "POST /shorten" {
handler = "PostShorten"
memory = var.memory
}
}
}
backbone {
nosql "links" {
size = var.store
seed = var.seed
}
}
edge {
rate_limit = var.rate
site_auth {
realm = "links"
paths = ["*"]
user "review" {
password = env("REVIEW_PASS")
}
}
}Here links is production, links-staging is a free slice with a smaller Function booking, a 16MiB collection, a lower rate limit and demo rows, and links-sandbox is a development slice that books less memory and keeps everything else.
The blocks
- attributes
default- meaning
- A value the file reads as
var.name. It can be a string, number, bool, list or map.
- attributes
vars,slice- meaning
- Another slice from the same file.
varsis a map of the variables it sets, by name;slicenames the slice it applies to.
A variable takes the environment's value from vars, else its default. A variable with neither is an error at its line. An environment that sets a variable no variable block declares is an error too. A value in vars is a string, and it is read as the number or bool the attribute needs, so n = "7" fills depth = var.n.
A variable stands for an attribute's value: a memory booking, a size, a rate limit, a seed path, a list of allowed addresses. A block's label, such as a Function's route or a collection's name, is written out.
Which slice an environment is
The name is derived, not declared:
| environment | applies to |
|---|---|
prod and production | the bare slice, here links |
any other, such as staging | slice-environment, here links-staging |
any, with slice = "..." | the slice it names, here links-sandbox for dev |
The slice name follows the name rules, so slice-environment is at most 32 characters. An environment the file does not declare is refused, naming the ones it does:
Driftfile
Driftfile:1:1: Unknown environment: there is no environment "nope": the file declares stagingWorking with an environment
| command | does |
|---|---|
drift plan staging | Compares the file, with staging's variables, to the links-staging slice, and quotes its price |
drift apply staging | Creates or resizes that slice, then deploys into it |
drift apply plan.json | Applies a plan file; it records the environment it was planned for |
drift file run staging | Builds and runs the staging project locally |
drift file test staging | Runs its test blocks against the local run |
drift file resolve --env staging | Prints the shape the file compiles to for staging: every booking, size and limit. Offline |
The plan says which slice it is looking at:
$ drift plan staging
· environment staging → slice links-stagingAn environment follows the same rule as any slice: its shape decides whether it can be one of your free slices. In the example staging is small enough to fit the free slice, so drift apply staging asks whether to make it one of your free slices; a collection of the production 64MiB would not fit, which is why store is a variable too. There is no separate flag for an environment. drift apply does not yet move an existing slice between free and priced; when a free slice's shape no longer fits it refuses, and drift slice resize makes that move.
Secrets per environment
Config lives in the Driftfile; values that differ per environment live in files beside it. env("NAME") reads a variable from the terminal, then from --var NAME=value, then from .env.environment (for staging, .env.staging), then from .env, in that order. The environment's own file wins over the base one, and --no-env-file skips both.
# .env.staging, beside the Driftfile
REVIEW_PASS=staging-review-passThe name inside env() is always written out, so the same name carries a different value in each environment. A secret, an API key and a site password are references only, in every environment.
Where an environment can differ
Everything a variable can reach differs per environment: the memory a Function books, a storage size, a limits figure, the rate limit, an edge allowlist, a seed file, an environment variable's value. The edge section describes what each of those settings does, and a variable can stand in any of them.
This is not a CI pipeline: there is no "deploy staging, test, promote prod" orchestration in the Driftfile. Drift gives the verbs; your CI chains them, with plan files.