Clean disposable CI stacks on interruption and job completion
creative workspace review checks / verify (pull_request) Failing after 7m18s

This commit is contained in:
2026-10-31 02:14:35 +08:00
parent 40c1023a86
commit a589f51698
5 changed files with 136 additions and 3 deletions
+22
View File
@@ -0,0 +1,22 @@
#!/usr/bin/env bash
set -Eeuo pipefail
: "${GITEA_RUN_ID:?A numeric CI run ID is required}"
[[ "$GITEA_RUN_ID" =~ ^[1-9][0-9]*$ ]] || { echo 'Refusing invalid CI run ID' >&2; exit 2; }
project="prompt-ci-$GITEA_RUN_ID"
filter="label=com.docker.compose.project=$project"
# Recheck ownership immediately before removal; never target a shared stack.
containers=$(docker ps -aq --filter "$filter")
for resource in $containers; do
[[ "$(docker inspect --format '{{index .Config.Labels "com.docker.compose.project"}}' "$resource")" == "$project" ]] || { echo "Refusing foreign resource: $resource" >&2; exit 2; }
docker rm -f "$resource"
done
volumes=$(docker volume ls -q --filter "$filter")
for resource in $volumes; do
[[ "$(docker volume inspect --format '{{index .Labels "com.docker.compose.project"}}' "$resource")" == "$project" ]] || { echo "Refusing foreign resource: $resource" >&2; exit 2; }
docker volume rm "$resource"
done
networks=$(docker network ls -q --filter "$filter")
for resource in $networks; do
[[ "$(docker network inspect --format '{{index .Labels "com.docker.compose.project"}}' "$resource")" == "$project" ]] || { echo "Refusing foreign resource: $resource" >&2; exit 2; }
docker network rm "$resource"
done
+14 -2
View File
@@ -2,7 +2,8 @@
set -Eeuo pipefail
: "${WATERFALL_WIKI_IMAGE:?Build the runtime image first}"
: "${WATERFALL_WIKI_TEST_IMAGE:?Build the integration image first}"
export COMPOSE_PROJECT_NAME="prompt-ci-${GITEA_RUN_ID:-$$}"
export GITEA_RUN_ID="${GITEA_RUN_ID:-$$}"
export COMPOSE_PROJECT_NAME="prompt-ci-$GITEA_RUN_ID"
export PG_PORT=0 S3_PORT=0 WEB_PORT=0 WEB_BIND_ADDRESS=127.0.0.1
TEST_ENV=$(mktemp)
chmod 0600 "$TEST_ENV"
@@ -13,8 +14,19 @@ if [[ -n "${STUDIO_REVIEW_IMAGE:-}" ]]; then
STUDIO_SERVICES+=(prompt-studio)
fi
compose(){ docker compose -p "$COMPOSE_PROJECT_NAME" --env-file "$TEST_ENV" "${COMPOSE_FILES[@]}" "$@"; }
cleanup(){ code=$?; if [ "$code" -ne 0 ]; then compose logs --tail 160 waterfall-wiki waterfall-wiki-worker "${STUDIO_SERVICES[@]}"; fi; compose down -v --remove-orphans; rm -f "$TEST_ENV"; }
cleanup(){
code=$?
trap - EXIT INT TERM
if [ "$code" -ne 0 ]; then compose logs --tail 160 waterfall-wiki waterfall-wiki-worker "${STUDIO_SERVICES[@]}" || true; fi
cleanup_code=0
bash scripts/cleanup-integration.sh || cleanup_code=$?
rm -f "$TEST_ENV"
if [ "$code" -ne 0 ]; then exit "$code"; fi
exit "$cleanup_code"
}
trap cleanup EXIT
trap 'exit 130' INT
trap 'exit 143' TERM
# Random credentials exist only for this disposable stack; no host Node required.
docker run --rm --entrypoint node "$WATERFALL_WIKI_IMAGE" -e "const c=require('crypto');const hex=()=>c.randomBytes(24).toString('hex');console.log('POSTGRES_PASSWORD='+hex()+'\nS3_ACCESS_KEY='+hex()+'\nS3_SECRET_KEY='+hex())" > "$TEST_ENV"
# Explicitly isolate legacy unauthenticated integration coverage; production fails closed.
+80
View File
@@ -0,0 +1,80 @@
#!/usr/bin/env python3
"""Exercise cleanup ownership checks without a Docker daemon or build."""
import json
import os
from pathlib import Path
import subprocess
import tempfile
import unittest
SCRIPT = Path(__file__).with_name('cleanup-integration.sh')
FAKE_DOCKER = '''#!/usr/bin/env python3
import json, os, sys
args = sys.argv[1:]
with open(os.environ['DOCKER_CALLS'], 'a') as output:
output.write(json.dumps(args) + '\\n')
if os.environ.get('EMPTY_INVENTORY'):
sys.exit(0)
if args[0] == 'ps':
if os.environ.get('FAIL_LIST'):
sys.exit(1)
print('test-container')
elif args[:2] == ['volume', 'ls']:
print('test-volume')
elif args[:2] == ['network', 'ls']:
print('test-network')
elif 'inspect' in args:
print(os.environ.get('RESOURCE_OWNER', 'prompt-ci-605'))
'''
class CleanupTests(unittest.TestCase):
def run_cleanup(self, run_id='605', **extra):
with tempfile.TemporaryDirectory() as directory:
root = Path(directory)
docker = root / 'docker'
docker.write_text(FAKE_DOCKER)
docker.chmod(0o700)
calls = root / 'calls'
env = {**os.environ, 'PATH': directory + os.pathsep + os.environ['PATH'],
'GITEA_RUN_ID': run_id, 'DOCKER_CALLS': str(calls), **extra}
result = subprocess.run(['bash', str(SCRIPT)], env=env, capture_output=True)
return result.returncode, [json.loads(line) for line in calls.read_text().splitlines()] if calls.exists() else []
def test_removes_only_run_owned_resources_in_dependency_order(self):
code, calls = self.run_cleanup()
self.assertEqual(code, 0)
self.assertEqual([args for args in calls if 'rm' in args], [
['rm', '-f', 'test-container'], ['volume', 'rm', 'test-volume'],
['network', 'rm', 'test-network']])
for args in calls:
if '--filter' in args:
self.assertEqual(args[-1], 'label=com.docker.compose.project=prompt-ci-605')
self.assertEqual(sum('inspect' in args for args in calls), 3)
def test_already_clean_stack_is_a_successful_noop(self):
code, calls = self.run_cleanup(EMPTY_INVENTORY='1')
self.assertEqual(code, 0)
self.assertEqual(len(calls), 3)
self.assertFalse(any('rm' in args for args in calls))
def test_invalid_or_production_project_never_contacts_docker(self):
for value in ['', 'prompt', '0', '-1', '605*', '605/../prompt']:
with self.subTest(value=value):
code, calls = self.run_cleanup(value)
self.assertNotEqual(code, 0)
self.assertEqual(calls, [])
def test_foreign_label_prevents_removal(self):
code, calls = self.run_cleanup(RESOURCE_OWNER='prompt')
self.assertNotEqual(code, 0)
self.assertFalse(any('rm' in args for args in calls))
def test_inventory_failure_does_not_continue_with_deletion(self):
code, calls = self.run_cleanup(FAIL_LIST='1')
self.assertNotEqual(code, 0)
self.assertEqual(len(calls), 1)
if __name__ == '__main__':
unittest.main()