Reset a seeded environment

Seeding is slow. Do it once, snapshot, and reset from the snapshot for the rest of the project's life.

Runs against the demo project.

1. Seed everything, once

Schema and reference data

docker compose exec -T postgres psql -U postgres -d demo <<'SQL'
CREATE TABLE tenants (id serial PRIMARY KEY, name text NOT NULL);
CREATE TABLE orders  (id serial PRIMARY KEY, tenant_id int REFERENCES tenants(id), total numeric);
INSERT INTO tenants (name) VALUES ('acme'), ('globex'), ('initech');
INSERT INTO orders (tenant_id, total) SELECT (n % 3) + 1, n * 1.5 FROM generate_series(1, 5000) n;
SQL

Uploaded files

docker compose exec -T minio sh -c '
  mkdir -p /data/demo/invoices
  for i in 1 2 3; do echo "invoice $i" > /data/demo/invoices/inv-$i.txt; done
'

A warm cache

docker compose exec -T redis redis-cli MSET \
  tenants:count 3 orders:count 5000 cache:warmed 1

2. Capture the baseline

arcave savepoint create clean --description "seeded baseline — do not delete"

stdout (trimmed)

{
  "id": "019e83ff-30c0-7000-8000-000000000000",
  "name": "clean",
  "description": "seeded baseline — do not delete",
  "created_at": "2026-06-01T18:36:40.002114Z",
  "stats": [
    { "volume": { "object_id": "volume-named-arcave-demo_pgdata",    "type": "named", "backend": "fullcopy" }, "type": "fullcopy" },
    { "volume": { "object_id": "volume-named-arcave-demo_miniodata", "type": "named", "backend": "fullcopy" }, "type": "fullcopy" },
    { "volume": { "object_id": "volume-anon-redis-1-data",          "type": "anon",  "backend": "fullcopy" }, "type": "fullcopy" }
  ],
  "writable_layers": [
    { "service": "minio",    "replica": 1, "container_name": "arcave-demo-minio-1",    "upstream_image_ref": "minio/minio:latest" },
    { "service": "postgres", "replica": 1, "container_name": "arcave-demo-postgres-1", "upstream_image_ref": "postgres:16" },
    { "service": "redis",    "replica": 1, "container_name": "arcave-demo-redis-1",    "upstream_image_ref": "redis:7" }
  ],
  "images": [
    { "id": "sha256:…", "refs": ["minio/minio:latest"], "stored": true },
    { "id": "sha256:…", "refs": ["postgres:16"],        "stored": true },
    { "id": "sha256:…", "refs": ["redis:7"],            "stored": true }
  ],
  "spec": { "format": "compose-yaml" }
}

Trimmed to the fields that identify each entry; Roll back a failed migration shows one of each in full. Three volumes, three writable layers, three images, one savepoint.

3. Let a test suite make a mess

docker compose exec -T postgres psql -U postgres -d demo -c 'TRUNCATE orders CASCADE;'
docker compose exec -T minio sh -c 'rm -rf /data/demo/invoices'
docker compose exec -T redis redis-cli FLUSHALL

4. Reset

arcave rollback clean

5. Verify all three came back

docker compose exec -T postgres psql -U postgres -d demo -tAc 'SELECT count(*) FROM orders;'

5000

docker compose exec -T minio sh -c 'ls /data/demo/invoices'

inv-1.txt
inv-2.txt
inv-3.txt

docker compose exec -T redis redis-cli MGET tenants:count orders:count cache:warmed

1) "3"
2) "5000"
3) "1"

This is the point of whole-project capture. A pg_restore would have given you the first of those three, and left the object store empty and the cache cold.

6. Wire it into the test suite

run-tests.sh

#!/usr/bin/env bash
set -euo pipefail

export ARCAVE_PROJECT_NAME=arcave-demo

arcave rollback clean          # reset BEFORE, not after
./integration-tests.sh

Resetting before rather than after has two advantages: a crashed suite cannot skip its own cleanup, and you can inspect the wreckage of a failed run before the next one wipes it.

7. Protect the baseline

arcave savepoint list

stdout (trimmed)

[
  { "id": "019e8488-3ac8-7000-8000-000000000000", "name": "after-refactor", "description": "", "created_at": "2026-06-01T21:13:01.114872Z" },
  { "id": "019e83ff-30c0-7000-8000-000000000000", "name": "clean", "description": "seeded baseline — do not delete", "created_at": "2026-06-01T18:36:40.002114Z" }
]

warning

arcave savepoint remove --retain-recent 5 will delete clean once five newer savepoints exist. Retention rules do not know which savepoint is precious.

Prune by id instead:

arcave savepoint remove 019e8488-3ac8-7000-8000-000000000000

Or keep the baseline in a session of its own. Retention works within one session, so no sweep in the session you work in can reach it:

arcave session create --name fixtures
arcave --session fixtures savepoint create clean --description "seeded baseline — do not delete"

run-tests.sh, resetting from the fixtures session

arcave --session fixtures rollback clean
./integration-tests.sh

Take the savepoint right after seeding, as in step 2. The commands that name no session keep working in yours, and arcave savepoint remove --retain-recent 5 there never sees clean.

Related docs

• Resetting fixtures and demos — the pattern.
• Prune savepoints with retention rules — the next example.
• savepoint remove — the retention rules in full.