Backing up Meilisearch: dumps vs snapshots, explained

Meilisearch ships two backup mechanisms: dumps (portable, version-flexible, slower) and snapshots (fast, binary, version-locked). Dumps are for migrations and disaster recovery across servers. Snapshots are for fast hot backups inside the same instance.

The 30-second answer

• Use a dump when migrating between servers, upgrading Meilisearch versions, or sharing data between environments.
• Use a snapshot when you want a fast hot backup of the current state (point-in-time recovery), restoring to the same version on the same machine.
• Run a snapshot daily and a dump weekly. Both are cheap.

Side-by-side comparison

Creating a dump

curl -X POST 'https://your-meilisearch/dumps' \
  -H 'Authorization: Bearer YOUR_ADMIN_KEY'

Returns immediately with a task envelope:

{
  "taskUid": 42,
  "indexUid": null,
  "status": "enqueued",
  "type": "dumpCreation",
  "enqueuedAt": "2026-04-11T12:00:00Z"
}

The dump file lands in the configured dumps directory (default ./data.ms/dumps/) with a name like 20260411-120000.dump. Watch the task via GET /tasks/{taskUid} until status flips to succeeded.

Restoring a dump

Restoring isn't an HTTP call — it's a Meilisearch startup flag:

meilisearch --import-dump ./dumps/20260411-120000.dump

The destination Meilisearch must be empty or you must use --ignore-dump-if-db-exists. The restore takes about as long as the original dump. Your indexes, documents, settings, and API keys are all reconstructed.

Creating a snapshot

curl -X POST 'https://your-meilisearch/snapshots' \
  -H 'Authorization: Bearer YOUR_ADMIN_KEY'

Same task-envelope shape as dumps. The snapshot file lands in the snapshots directory as {timestamp}.snapshot.

Restoring a snapshot

Same shape:

meilisearch --import-snapshot ./snapshots/20260411-120000.snapshot

Snapshots are version-locked — the restore must be on the exact same Meilisearch version that produced them. Trying to restore a 1.10 snapshot on a 1.11 server will fail.

Auto-generating snapshots on a schedule

Meilisearch can generate snapshots automatically without any external scheduler. Pass a flag at startup:

meilisearch \
  --schedule-snapshot 86400 \
  --snapshot-dir /var/lib/meilisearch/snapshots

This snapshots once per day (86400 seconds). Older snapshots are NOT auto-deleted — you need a cron job or filesystem rotation policy to keep disk usage in check.

What's actually inside a dump?

A dump file is a tarball containing JSON for each index, the index settings, the API keys, and metadata about the Meilisearch version that produced it. You can inspect it:

tar -tvf ./dumps/20260411-120000.dump
# instances.json
# indexes/movies/documents.jsonl
# indexes/movies/settings.json
# keys.jsonl
# metadata.json

What's inside a snapshot?

A binary copy of Meilisearch's on-disk LMDB database. You can't inspect it without loading it into Meilisearch — that's why it's faster (no serialization) but version-locked (the on-disk format isn't stable across versions).

Backup strategy by team size

Solo developer / small team

• Snapshot daily (auto-scheduled via --schedule-snapshot 86400)
• Dump weekly (manually triggered or via cron)
• Copy both off-server to S3 / Backblaze / wherever you keep other backups
• Keep the last 7 days of snapshots and the last 4 weeks of dumps

Production at scale

• Snapshot every 4–6 hours
• Dump nightly
• Off-site replication via your existing backup pipeline (rsync, S3 sync, etc.)
• Test restoration monthly — a backup you've never restored is not a backup

FAQ

Can I take a snapshot while users are searching?

Yes. Snapshots are non-blocking — Meilisearch creates a consistent point-in-time copy without pausing search traffic. Same for dumps, though dumps take longer.

Can I restore a dump from Meilisearch 1.6 onto Meilisearch 1.13?

Yes — dumps are forward-compatible. Restoring an older dump on a newer version is the supported upgrade path. Restoring a newer dump on an older version is NOT supported.

Where do dumps and snapshots get stored if I don't configure a directory?

Inside the Meilisearch data directory: ./data.ms/dumps/ and ./data.ms/snapshots/. If you're running in Docker with a mounted volume, both directories live inside the volume.

Can AdminDex trigger backups?

Yes — AdminDex's instance dashboard has a "Create dump" and "Create snapshot" button that triggers the same async task and shows the result in the live task viewer. No need to remember the curl invocation.