feat: comprehensive health check with DB, scheduler, cache

- Add /api/v1/health endpoint with component-level checks
- Verify DB connectivity, fail2ban socket, scheduler, session cache
- Add SQLite WAL cleanup on startup (orphan crash files)
- Migration 8: import_log.timestamp → INTEGER UNIX epoch
- Align import_log timestamps with history_archive (already UNIX int)
- Add unit tests for DB cleanup and health router

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
This commit is contained in:
2026-05-02 23:03:57 +02:00
parent b631c1c546
commit 1285bc8571
12 changed files with 472 additions and 241 deletions

View File

@@ -1,207 +1,3 @@
### Issue #6: HIGH - Inconsistent API Response Format Between Endpoints
**Where found**:
- `backend/app/routers/bans.py` - Returns `CommandResponse(success: bool, message: str)`
- `backend/app/routers/history.py` - Returns `HistoryListResponse(bans: [], pagination: {})`
- `backend/app/routers/status.py` - Returns `dict[str, str]` with no consistent structure
- Various routers return different error formats
**Why this is needed**:
Frontend cannot assume consistent structure, must write different parsing code for each endpoint. Type safety is compromised. Mapping backend schema to frontend types becomes fragile.
**Goal**:
Standardize on a unified response format across all endpoints to ensure type safety and consistency.
**What to do**:
1. Define standard response format:
```python
@dataclass
class ApiResponse(Generic[T]):
success: bool
data: T | None
error: str | None
error_code: str | None = None
pagination: Pagination | None = None
metadata: dict[str, Any] | None = None
```
2. Update all routers to use this format
3. Map endpoint-specific responses into `data` field
4. Add error codes for different error types
5. Update frontend to parse this format
6. Add validation that all endpoints conform to format
**Possible traps and issues**:
- Backward compatibility breaks (clients expecting old format fail)
- Existing frontend code expects old response format
- Performance impact of additional wrapping (measure and optimize)
- Migration burden - all routers need updates
**Docs changes needed**:
- Update API documentation with new response format
- Add examples of standard response across different scenarios
- Create migration guide for API consumers
**Doc references**:
- DATABASE_API_DEPLOYMENT_ISSUES.md - Issue "2.1 Inconsistent Response Format"
---
### Issue #7: HIGH - Docker Health Check Fails in Production
**Where found**:
- `Docker/compose.prod.yml` (lines 40-47)
- Backend health check runs Python in subprocess but image has no Python in runtime
- Frontend health check uses `wget` that may not exist
**Why this is needed**:
Health checks always fail, container marked unhealthy, Kubernetes/orchestrators evict pod thinking service is down. Cascades to service outages.
**Goal**:
Implement reliable health checks that actually verify service health without depending on missing tools.
**What to do**:
1. Replace Python subprocess check with simple HTTP request:
```yaml
healthcheck:
test: ["CMD", "curl", "-f", "http://localhost:8000/api/health"]
```
2. Implement comprehensive health check endpoint that verifies:
- Database is accessible
- fail2ban socket is reachable
- Background scheduler is healthy
- Cache systems are initialized
3. Return 503 if any component is unhealthy
4. Add timeout and retry parameters:
```yaml
interval: 30s
timeout: 10s
retries: 3
start_period: 40s
```
5. Test health check in both Alpine and full Python images
**Possible traps and issues**:
- curl might not be installed in lightweight images (use `wget` as fallback or multi-tool)
- Health check runs frequently (every 30s) - must be lightweight
- False positives if timeout is too short
- Can't check fail2ban socket during initial startup (timing issue)
**Docs changes needed**:
- Add troubleshooting section to `Docs/TROUBLESHOOTING.md` - "502 Bad Gateway errors"
- Document health check behavior in `Docs/Deployment.md`
**Doc references**:
- DATABASE_API_DEPLOYMENT_ISSUES.md - Issue "6.2 Missing Health Checks in Production"
- `Docker/compose.prod.yml` - Current health check definition
---
### Issue #8: HIGH - Database Migration Fragility (Rollback Missing)
**Where found**:
- `backend/app/startup.py` (lines 105-130) - `run_migrations()`
- `backend/app/db.py` (lines 132-142) - Schema migration table
- No transaction wrapping migrations
**Why this is needed**:
If a migration fails mid-transaction, the database schema becomes inconsistent:
- Some tables created, others not
- Next startup tries to re-apply migration
- Second attempt fails because column already exists
- Application can't start
**Goal**:
Implement transactional migrations with automatic rollback on failure.
**What to do**:
1. Wrap each migration in explicit transaction:
```python
async def run_migration(name: str, migration_func):
try:
await db.execute("BEGIN EXCLUSIVE")
await migration_func()
await db.execute(
"INSERT INTO schema_migrations (name, applied_at) VALUES (?, ?)",
name, int(time.time())
)
await db.execute("COMMIT")
except Exception:
await db.execute("ROLLBACK")
raise
```
2. Add idempotent migration checks (CREATE TABLE IF NOT EXISTS)
3. Implement downtime procedure for failed migrations
4. Add migration rollback capability
5. Test migration failures and recovery
**Possible traps and issues**:
- SQLite WAL mode can leave orphaned `.wal` files after crashes (handle cleanup)
- Long-running migrations might timeout
- Complex migrations spanning multiple steps hard to make atomic
- Rollback procedure must be well-documented and tested
- Development vs production migration paths might diverge
**Docs changes needed**:
- Create `Docs/DATABASE_MIGRATIONS.md` explaining migration process
- Add disaster recovery section to `Docs/Deployment.md`
- Document in `Docs/TROUBLESHOOTING.md` - "Database migration failed"
**Doc references**:
- DATABASE_API_DEPLOYMENT_ISSUES.md - Issue "6.3 Database Migration Issues"
- `backend/app/startup.py` - Migration runner
---
### Issue #9: HIGH - No Graceful Shutdown Handling
**Where found**:
- `Docker/Dockerfile.backend` (lines 75-76) - Entrypoint doesn't handle signals
- `backend/app/main.py` - Lifespan doesn't gracefully shutdown
- No SIGTERM handler in background tasks
- Container receives SIGKILL without cleanup opportunity
**Why this is needed**:
Without graceful shutdown:
- In-flight ban requests don't complete
- Background jobs interrupted mid-import
- Incomplete blocklist imports leave stale data
- Database connections don't close properly
**Goal**:
Implement graceful shutdown that allows pending operations to complete before process exits.
**What to do**:
1. Implement lifespan context manager that handles shutdown:
```python
@asynccontextmanager
async def lifespan(app: FastAPI):
# Startup
yield
# Shutdown: allow pending tasks
tasks = [t for t in asyncio.all_tasks() if not t.done()]
await asyncio.gather(*tasks, return_exceptions=True)
```
2. Set graceful_timeout in Dockerfile: `docker stop --time=30`
3. Handle SIGTERM signal to trigger shutdown
4. Drain in-flight requests before exit
5. Close database connections and scheduler cleanly
6. Add logging for shutdown events
**Possible traps and issues**:
- Tasks might take longer than shutdown timeout (configure appropriate timeout)
- Hanging tasks won't terminate (need proper cancellation)
- Scheduler might reject new jobs during shutdown
- Race conditions between shutdown and new requests
**Docs changes needed**:
- Add deployment best practices to `Docs/Deployment.md`
- Document graceful shutdown in `Docs/TROUBLESHOOTING.md`
**Doc references**:
- DATABASE_API_DEPLOYMENT_ISSUES.md - Issue "6.4 No Graceful Shutdown"
---
### Issue #10: HIGH - Database Type Inconsistency (Timestamps Mixed Across Tables)
**Where found**: