Hot Reload

Rindexer supports hot-reloading your rindexer.yaml configuration without manually stopping and restarting the process. When enabled with the --watch flag, rindexer monitors your YAML file for changes, validates the new configuration, and automatically restarts with the updated settings.

Quick Start

Add the --watch flag (or -w) before the subcommand:

rindexer start --watch all

Now edit your rindexer.yaml — rindexer will automatically detect the change, validate the new config, and restart.

How It Works

When --watch is enabled, rindexer runs as two processes:

1. Outer process — A lightweight restart loop that spawns and monitors the indexer
2. Inner process — The actual indexer with a file watcher attached

The inner process watches rindexer.yaml using OS-native file events (FSEvents on macOS, inotify on Linux). When a change is detected:

1. Debounce — Waits 500ms for rapid successive saves to settle
2. Validate — Parses the new YAML. If invalid, the change is rejected and the current config keeps running
3. Diff — Compares old and new manifests to classify the change
4. Graceful shutdown — Stops the GraphQL server, health server, and active indexing tasks
5. Restart — The process exits with code 75, the outer loop catches it and spawns a fresh process

Change Classification

Not all changes are treated equally. Rindexer computes a diff to determine the appropriate action:

Error Handling

If you save an invalid rindexer.yaml, rindexer will log an error and keep the current configuration running:

ERROR Hot-reload: new manifest is invalid, keeping current config: ...

Fix the YAML error and save again — rindexer will pick up the corrected file.

Production Deployment

The --watch flag includes a built-in restart loop that works out of the box for local development. For production, you can either use the built-in loop or rely on your process manager to handle restarts.

Docker

services:
  rindexer:
    image: your-rindexer-image
    command: rindexer start --watch all
    restart: unless-stopped

The container will restart automatically when rindexer exits with code 75 on config change. It will stay stopped on clean shutdown (exit code 0) or if you run docker compose down.

systemd

[Unit]
Description=rindexer indexer
After=network.target postgresql.service
 
[Service]
ExecStart=/usr/local/bin/rindexer start --watch all
WorkingDirectory=/path/to/your/project
Restart=on-failure
RestartSec=1
Environment=DATABASE_URL=postgresql://user:pass@localhost/db
 
[Install]
WantedBy=multi-user.target

With Restart=on-failure, systemd treats exit code 75 as a failure and restarts the service. Exit code 0 (clean shutdown via Ctrl+C or SIGTERM) is treated as success and stops the service.

Kubernetes

apiVersion: apps/v1
kind: Deployment
spec:
  template:
    spec:
      containers:
        - name: rindexer
          command: ["rindexer", "start", "--watch", "all"]
          # Kubernetes restarts containers automatically on non-zero exit
          # Mount your rindexer.yaml via ConfigMap for easy updates
          volumeMounts:
            - name: config
              mountPath: /app/rindexer.yaml
              subPath: rindexer.yaml
      volumes:
        - name: config
          configMap:
            name: rindexer-config

Update the ConfigMap and rindexer will detect the change and restart:

kubectl create configmap rindexer-config --from-file=rindexer.yaml -o yaml --dry-run=client | kubectl apply -f -

Limitations

• No-code projects only — Rust projects require recompilation, which is outside the scope of hot reload
• Full process restart — Each reload restarts the entire indexer process. This means historical indexing re-runs from the last checkpoint (no data is lost, but there is a brief pause in live indexing)
• Exit code 75 — The process exits with code 75 (EX_TEMPFAIL) to signal a restart. Make sure your monitoring does not treat this as a crash