Dumps

Meilisearch stores its database in files located in ./data.ms by default. Because this database is bound to the version of Meilisearch that created it, you must use dumps to migrate data between Meilisearch releases.

A dump is a compressed file containing an export of your Meilisearch instance. It contains all your indexes, documents, and settings, but in a raw unprocessed form. A dump isn't an exact copy of your database—it is closer to a blueprint that allows you to create an identical dataset.

Creating a dump is also referred to as exporting it, whereas launching Meilisearch with a dump is referred to as importing it.

To create a dump of your dataset, use the create a dump endpoint:

curl \
  -X POST 'http://localhost:7700/dumps'

The above code triggers a dump creation process. It also returns a summarized task object that you can use to check the status of your dump.

{
  "taskUid": 1,
  "indexUid": null,
  "status": "enqueued",
  "type": "dumpCreation",
  "enqueuedAt": "2022-06-21T16:10:29.217688Z"
}

In the below command, replace 1 with the taskUid returned by the previous command.

curl \
  -X GET 'http://localhost:7700/tasks/1'

This command should return an object with detailed information about the dump operation:

{
  "uid": 1,
  "indexUid": null,
  "status": "succeeded",
  "type": "dumpCreation",
  "canceledBy": null,
  "details": {
    "dumpUid": "20220621-161029217"
  },
  "error": null,
  "duration": "PT0.025872S",
  "enqueuedAt": "2022-06-21T16:10:29.217688Z",
  "startedAt": "2022-06-21T16:10:29.218297Z",
  "finishedAt": "2022-06-21T16:10:29.244169Z"
}

The dump creation process is an asynchronous task that takes time proportional to the size of your dataset. All indexes of the current instance are exported along with their documents and settings and saved as a single .dump file.

After dump creation is finished—when status is succeeded—the dump file is added to the dump directory. By default, this folder is named dumps and can be found in the same directory as your Meilisearch binary. You can customize this using the --dump-dir configuration option. If the dump directory does not already exist when the dump creation process is called, Meilisearch will create it.

If a dump file is visible in the file system, the dump process was successfully completed. Meilisearch will never create a partial dump file, even if you interrupt an instance while it is generating a dump.

Import dumps by launching a Meilisearch instance with the --import-dump configuration option.

During a dump import, all indexes contained in the indicated .dump file are imported along with their associated documents and settings. Any existing index with the same uid as an index in the dump file will be overwritten.

While a dump is being imported, the API is not available to the task queue. As a result, no read or write operations can be performed until the importing process is complete.

Once you have exported a dump you can use the resulting .dump file to launch Meilisearch with the --import-dump configuration option.

As the data contained in the dump needs to be indexed, the process will take some time to complete. Only when the dump has been fully imported will the Meilisearch server start, after which you can begin searching through your data.

./meilisearch --import-dump /dumps/20200813-042312213.dump

Dumps are used to restore your database after updating Meilisearch or to copy your database to other Meilisearch instances without having to worry about their respective versions. For more on this subject, see a comparison of snapshots and dumps.