Speed Up Typesense Startup in Production with Disk Persistence

typesense
Published

May 10, 2025

If you’re running Typesense in production with large collections or semantic search, startup time can become a bottleneck. By default, Typesense rebuilds its in-memory index from scratch on each boot — which means loading and re-indexing every document. This can take minutes (or hours) as your dataset grows.

To fix this, enable disk persistence and snapshots so your data and indexes are saved between restarts.

🔧 How to Enable Disk Persistence

Edit your typesense-server.ini:

enable_disk_persistence = true
snapshot_interval_seconds = 300

This configuration tells Typesense to:

  • Save a snapshot of in-memory state every 5 minutes.
  • Reuse that snapshot to skip rebuilding indexes on startup.

Then restart the server:

sudo systemctl restart typesense-server

⸻

✅ Verifying It Works

You can confirm snapshot saving and reuse in two ways:

  1. Log Messages

Run:

journalctl -u typesense-server -f

Look for messages like:

Saving snapshot for collection content_en Restoring collection from snapshot

This confirms that the collection was successfully persisted and restored.

  1. Check Snapshot Files on Disk

Navigate to your data-dir, e.g.:

cd /mnt/data/typesense/data/

There should be a folder of .sst file that looks something like this:

state/snapshot/snapshot_00000000000000394299/db_snapshot/

These files store the serialized in-memory state and write-ahead log.

⸻

🚀 Result: Faster Restarts

Without disk persistence, Typesense must reload and re-index every document from disk. With snapshots enabled, startup time drops dramatically — even with millions of documents.

Documents Without Persistence With Persistence
100k ~2 minutes ~5 seconds
1M ~20 minutes ~15 seconds

⸻

🛡️ Production-Ready Tip

Enable disk persistence for any production deployment, especially when using:

  • Large text collections
  • ONNX/embedding-based semantic search
  • Limited system memory or shared environments

Snapshots keep your data safe and your APIs fast after restarts or crashes.

⸻