df-memgraph

name	df-memgraph
description	Manage the test Memgraph container (start/stop/restart/status/logs/clear/wipe) via docker-compose.test.yml. Use when graph tests need the DB up, data cleared, a stuck lock cleared, or the volume fully wiped.
disable-model-invocation	true
allowed-tools	Bash(docker compose *), Bash(poetry run python *)

Usage: /df-memgraph [start|stop|restart|status|logs|clear|wipe] (default: status)

Controls the test Memgraph + Memgraph Lab containers defined in docker-compose.test.yml. Run from the repo root. Graph tests require Memgraph running; Memgraph is a test-only dependency, never a production one.

Pick the command matching the argument. If no argument is given, run status.

start — bring the DB up

Starts Memgraph (bolt on 127.0.0.1:7687) and Memgraph Lab (http://localhost:3000). Idempotent — safe to run when already up.

docker compose -f docker-compose.test.yml up -d

stop — stop the containers, keep data

Stops and removes the containers but keeps the mg_lib volume, so data survives.

docker compose -f docker-compose.test.yml down

restart — clear a stuck lock, keep data

Use this when a pytest run was pkill -9'd mid-test and the autouse cleanup_graph_db fixture's DETACH DELETE lock is stuck (symptom: graph tests hang/deadlock). Restart clears the lock but does not clear the data (the mg_lib volume persists).

docker compose -f docker-compose.test.yml restart

status — is it up? (default)

docker compose -f docker-compose.test.yml ps

logs — recent Memgraph logs

docker compose -f docker-compose.test.yml logs --tail 80 memgraph

clear — wipe data only, container stays up

Fastest reset: clears all nodes, relationships, indexes, and constraints via cypher while the container keeps running (no volume destroy, no restart). Use during development when Memgraph gets polluted with test data. Requires the DB to be up.

poetry run python -c "
from gqlalchemy import Memgraph

db = Memgraph(host='127.0.0.1', port=7687)

# Drop all constraints first
print('Dropping constraints...')
for row in db.execute_and_fetch('SHOW CONSTRAINT INFO'):
    for prop in row['properties']:
        query = f\"DROP CONSTRAINT ON (n:{row['label']}) ASSERT n.{prop} IS UNIQUE\"
        print(f'  {query}')
        db.execute(query)

# Drop all indexes
print('Dropping indexes...')
for row in db.execute_and_fetch('SHOW INDEX INFO'):
    for prop in row['property']:
        query = f\"DROP INDEX ON :{row['label']}({prop})\"
        print(f'  {query}')
        db.execute(query)

# Delete all nodes and relationships
print('Deleting all nodes and relationships...')
result = list(db.execute_and_fetch('MATCH (n) DETACH DELETE n RETURN count(n) as deleted'))
deleted = result[0]['deleted'] if result else 0
print(f'  Deleted {deleted} nodes')

print()
print('Database reset complete.')
"

wipe — destroy the volume and start fresh

Full reset: tears down the containers and deletes the persistent mg_lib volume (all data gone), then brings a clean DB back up. Use when stale leftover nodes cause spurious failures in unrelated tests and a plain clear isn't enough (e.g. the stuck lock persists or the container itself is wedged).

docker compose -f docker-compose.test.yml down -v && docker compose -f docker-compose.test.yml up -d

---

Notes:

• Memgraph Lab (visual browser) is at http://localhost:3000 once start completes.
• Only one graph-test run at a time — concurrent pytest processes against the same Memgraph deadlock on the cleanup fixture.
• Reset ladder, cheapest first: clear (data only, container up) → restart (clears a stuck lock, keeps data) → wipe (destroys the volume, fresh DB).