Docs Menu
Docs Home
/ /
Atlas App Services
/ /

Sync Settings

On this page

  • Available Settings
  • Sync Type
  • Development Mode
  • Cluster to Sync
  • Database Name (Development Mode Only)
  • Queryable Fields
  • Permissions
  • Data Ingest
  • Client Max Offline Time
  • Client Recovery
  • Sync Configuration File Reference
  • Sync Config Object

This page explains the settings available when you enable or configure Device Sync.

Atlas Device Sync has two sync modes: Flexible Sync and the older Partition-Based Sync. Partition-Based Sync has been deprecated and is disallowed for new Sync configurations. If you have an existing app that uses Partition-Based Sync, you can migrate to Flexible Sync. For more information, refer to Migrate Device Sync Modes.

Flexible Sync lets you define a query in the client and sync only the objects that match the query. With client-side subscriptions, client applications can:

  • Maintain queries

  • React to changes

  • Add, change, or delete queries

Development Mode is a configuration setting that allows Device Sync to infer and update schemas based on client-side data models. This streamlines development, but should not be used in production.

Development Mode speeds up development by allowing you to design schemas directly in client application code.

When you sync a realm, Atlas App Services maps every synced object type to its own collection in the database specified by Database Name (Development Mode Only). If you update the object model on the client side, App Services updates the collection schema to match. This lets you update objects in your client code as you develop your app.

You can use data access rules with Development Mode. Note that schema changes ignore data access rules. This means that any client can update the backend schema by changing the client model.

To learn more about how the Realm Object Schemas map to the App Services Schemas when using Development Mode, refer to Data Model Mapping.

For more information about modifying synced object schemas, see: Update Your Data Model.


Disable Development Mode for Production Apps

Development mode is a development utility that is not suitable for production use. Make sure that you turn off Development Mode before you make your app accessible in a production environment.

App Services Apps in Development Mode that were created after September 13, 2023 can make breaking changes to synced object schemas from client code.

If your App was created before September 13, 2023, you can contact support to enable this feature.


  • App Services App created after September 13, 2023

  • MongoDB 5.0 or later for Flexible Sync compatibility

  • Minimum SDK version:

    • Realm C++ SDK v1.0.0

    • Realm Flutter SDK v1.6.0

    • Realm Java SDK v10.16.2

    • Realm Kotlin SDK v11.1.1

    • Realm .NET SDK v11.6.0

    • Realm Node.js SDK v12.2.0

    • Realm React Native SDK v12.2.0

    • Realm Swift SDK v10.42.2


Apps created before September 13, 2023

For Apps created before September 13, 2023, you need to update your object schema in the App Services UI. For details, see Update Your Data Model

To make a breaking change from client code:

  1. Delete your local realm and data. This will not affect data synced to the backend. Local changes that have not been synced will be removed and are unrecoverable.

  2. Change your local object model.

  3. Open a realm with your updated object model.

  4. Run your client app to sync your changes to the backend.

To delete a Realm file use the Realm SDK-specific methods:

Enabling Development Mode has two side effects:

If your app does not need anonymous authentication, you may want to disable it after enabling Development Mode.

You cannot re-enable deployment drafts in the UI until Development Mode is disabled. However, you can still manually create deployment drafts through the CLI or the Admin API.

The name of the Atlas cluster data source where you want the synced data stored.

Flexible Sync requires MongoDB 5.0. You cannot use Sync with a serverless instance. In the UI, clusters display as gray and not-selectable when they do not meet the requirements for Flexible Sync.

When you enable Development Mode, you specify a database to store synced objects. App Services creates new collections in this Development Mode database for every type of synced object.


Specify a Development Mode database of myapp. Your iOS client has a Person model. You sync a realm that contains an instance of the Person object. Development Mode creates a server-side schema associated with the model. The object syncs to the myapp.Person collection.

App Services continues creating new server-side schemas and collections for each new object type. If you later add a Dog object, that object will sync to a new myapp.Dog collection that App Services will create.

When you configure Flexible Sync, you specify field names that your client application can query in a Flexible Sync subscription. Fields that can be used in a subscription query are called queryable fields.


In a to-do list app, you might set assignee or owner as queryable fields. On the client side, you can then query for tasks whose assignee or owner matches the logged-in user.

Queryable fields apply to a scope you designate when you configure them. The two available scopes are:

  • Global queryable fields: scoped across all collections in an App's Schema.

  • Collection queryable fields: scoped to a single collection in the App.

Scoping a queryable field to a specific collection reduces the amount of backing Atlas storage required to store Sync metadata.

You can use rules and permissions to configure more granular access control on a per-collection basis. You can define collection-level rules and permissions for both global and collection queryable fields.

You can automatically specify queryable fields by enabling Development Mode. Fields that appear in client queries while using Development mode are automatically added as collection queryable fields for the collection being queried.

The field names you provide are arbitrary strings. If an object type has a field whose name matches a field name you provided (and meets other eligibility criteria), that field becomes available to Device Sync to query.

You may only add or remove an indexed queryable field when Device Sync is not enabled. If Device Sync is already running in your App, you must terminate Sync, and configure the indexed queryable field when you re-enable it.

This causes client resets for any client attempting to reconnect after re-enabling Sync.

Flexible Sync only supports top-level primitive fields with a scalar type as queryable fields. You can also include arrays of these primitives as queryable fields. Flexible Sync does not support embedded objects or arrays of objects as queryable fields.

Indexed queryable fields support a subset of data types. Your indexed queryable field can be one of: int64, string, ObjectId, UUID.


See also: Realm Query Language - Flexible Sync Limitations

For information on the queries you can perform on these fields, see: Flexible Sync RQL Limitations

App Services reserves some keywords for the Realm Query Language and other purposes. You cannot use reserved keywords as field names.

App Services reserves the following keywords with any capitalization:

  • and

  • asc

  • ascending

  • beginswith

  • between

  • contains

  • desc

  • descending

  • distinct

  • endswith

  • falsepredicate

  • inf

  • infinity

  • like

  • limit

  • nan

  • nil

  • null

  • or

  • sort

  • subquery

  • truepredicate


You cannot use descending, Descending, DESCENDING, or DeScEnDiNG as a field name.

App Services also reserves the following keywords with the given exact capitalization:

  • ALL

  • ANY

  • B64


  • IN

  • NONE

  • NOT

  • SOME

  • TRUE

  • all

  • any

  • false

  • in

  • none

  • not

  • oid

  • some

  • true

  • uuid


You cannot use true or TRUE, since both capitalizations are specifically reserved, but you can use True or tRUE as a field name.

Each queryable field adds additional metadata storage to your Atlas cluster and may lead to degraded write performance. You should have as few queryable fields as needed by your application, and scope them to the minimum number of collections required.

Many apps find a good balance between storage usage and query flexibility with at most 10 queryable fields applying to any single collection. For example, if you have 3 global queryable fields and 7 collection queryable fields, you have 10 queryable fields that apply to the collection.

If you have a field that you only want to query in one collection, but it is configured as a global queryable field, this unnecessarily consumes Atlas storage space. For example, if you have a user field in every collection, but you only use it for Sync queries in one collection, scoping that as a collection queryable field reduces storage requirements. Reducing the scope means that Sync does not have to maintain metadata for that field for the other collections where you are not querying on the user field.

If you need to reduce storage usage or improve performance, you can remove unneeded queryable fields from your App. However, be aware of the consequences of adding or removing queryable fields. For more information, refer to Consequences of Adding or Removing Queryable Fields.

For additional considerations, refer to optimizing performance and storage when using Flexible Sync.

You can improve performance for certain types of workloads by adding an indexed queryable field. An indexed queryable field is a global queryable field that can be queried on more efficiently, providing improved Sync performance. You can designate one global queryable field as an indexed queryable field.

Indexing a queryable field improves performance for simple queries on a single field, such as {“store_id”: 1} or {“user_id”: “641374b03725038381d2e1fb”}.

The indexed queryable field must appear in the schemas of all of your Sync collections, and it must use the same valid data type. For example, if your indexed queryable field is store_id, it must appear in all of the collections you sync, and it must be the same valid type in all the collections. For more information about eligible field types, refer to Eligible Field Types.

You Can't Change Indexed Queryable Field Values on the Client

After you configure an indexed queryable field, client devices cannot update an existing object's indexed queryable field value. For example, if your indexed queryable field is store_id, the client cannot change this value directly. Changing it from the client is not supported because it may conflict with other updates made to the object in the same timeframe.

If you attempt to change an indexed queryable field's value on the device, this triggers a compensating write error. For more information about this error and the behavior it entails, refer to ErrorCompensatingWrite in the Flexible Sync Errors documentation.

You can still change this value directly in the Atlas database.


Changing an object's indexed queryable field value through Atlas may overwrite concurrent client updates to the object.

Client-Side Queries on Indexed Queryable Fields

When your app uses an indexed queryable field, client-side queries in a Flexible Sync subscription must include the indexed queryable field using an == or IN comparison against a constant at least once. For example, user_id == 641374b03725038381d2e1fb or store_id IN {1,2,3}.

You can optionally include an AND comparison as long as the indexed queryable field is directly compared against a constant using == or IN at least once. For example, store_id IN {1,2,3} AND region=="Northeast" or store_id == 1 AND (active_promotions < 5 OR num_employees < 10).

Invalid Flexible Sync queries on an indexed queryable field include queries where:

  • The indexed queryable field does not use AND with the rest of the query. For example store_id IN {1,2,3} OR region=="Northeast" is invalid because it uses OR instead of AND. Similarly, store_id == 1 AND active_promotions < 5 OR num_employees < 10 is invalid because the AND only applies to the term next to it, not the entire query.

  • The indexed queryable field is not used in an equality operator. For example store_id > 2 AND region=="Northeast" is invalid because it uses only the > operator with the indexed queryable field and does not have an equality comparison.

  • The query is missing the indexed queryable field entirely. For example, region=="Northeast" or truepredicate are invalid because they do not contain the indexed queryable field.

You can update your Sync configuration to add or remove queryable field names while Sync is enabled, but be aware of the following:

When you add a queryable field, devices can only sync on that field once the device has caught up to the point in time in Device Sync History where the field was added.

When you remove a queryable field, any devices still using that field will have their Device Sync session dropped and must perform a client reset. Clients not using the removed field won't receive any errors. To avoid triggering a client reset when you remove the queryable field, you should first remove usage of that field on the client-side.

If you terminate Sync before adding or removing queryable fields, these considerations do not apply. However, terminating Sync does trigger a client reset for any client that has Synced with your App.

Atlas Device Sync enforces role-based data access rules for all requests to a synced cluster. Rules are dynamic JSON expressions that determine a user's ability to sync, view, and modify data.

For details, see Role-based Permissions.

Data Ingest is a sync strategy for applications with heavy client-side insert-only workloads. You can enable it for one or more collections. It supports writing to any collection type, including an Atlas time-series collection.

For example, an IoT app that frequently logs sensor data has a significant write workload and no read workload. The device may also be offline for extended periods of time. Data Ingest bypasses some of the processing required for bi-directional sync, significantly improving write speed to an Atlas collection.

Other use cases include writing immutable data, such as invoices from a retail app, or logging application events, neither of which requires conflict resolution.

You can apply Data Ingest to individual collections. This means your app can use Data Ingest to write some data, but bi-directional Flexible Sync on other collections.

Data Ingest collections are only for writing data. You cannot use Flexible Sync queries against these collections. Instead, use Connect to MongoDB Data Sources.

After you have enabled Data Ingest, you implement it in the client app via the client SDKs. Currently, the following Realm SDKs support Data Ingest:

Atlas Device Sync completely manages the lifecycle of this data. It is maintained on the device until Data Ingest synchronization is complete, and then removed from the device.

Client Maximum Offline Time determines how long the client can be offline between sync sessions. Changing this value enables you to balance offline access with storage used in the synced Atlas cluster. For more information, refer to Client Maximum Offline Time.

Client Recovery enables the client to attempt to automatically perform a client reset while recovering data on the device. For more information, refer to Recover Unsynced Changes.

You can find the Sync configuration file for your application in the sync directory of an exported app:

└── sync/
└── config.json

For example, the following Sync configuration applies to apps using Flexible Sync.

"type": "flexible",
"development_mode_enabled": <Boolean>,
"service_name": "<Data Source Name>",
"database_name": "<Development Mode Database Name>",
"state": <"enabled" | "disabled">,
"client_max_offline_days": <Number>,
"is_recovery_mode_disabled": <Boolean>,
"queryable_fields_names": [
<Array of String Field Names>
"indexed_queryable_fields_names": [
<Array of String Field Names>
"collection_queryable_fields_names": <Map[String][]String>
"permissions": "<Deprecated, Do Not Use>"

The deprecated permissions field might still appear in your exported app's configuration. That might indicate your app has not automatically migrated to the unified rule system yet. Please avoid deleting this field until your app has been migrated.

"type": "flexible",
"development_mode_enabled": <Boolean>,
"service_name": "<Data Source Name>",
"database_name": "<Development Mode Database Name>",
"state": <"enabled" | "disabled">,
"client_max_offline_days": <Number>,
"is_recovery_mode_disabled": <Boolean>,
"queryable_fields_names": ["<Field Name>", ...],
"indexed_queryable_fields_names": ["<Field Name>", ...],
"collection_queryable_fields_names": {
"<Collection Name>": ["<Field Name>", ...],

The sync mode. There are two Sync modes: Flexible Sync and the older Partition-Based Sync. We recommend using Flexible Sync. For more information about Partition-Based Sync, refer to Partition-Based Sync.

Valid Options for a Flexible Sync Configuration:

  • "flexible"

If true, Development Mode is enabled for the application. While enabled, App Services automatically stores synced objects in a specific database (specified in database_name) and mirrors objects types in that database's collection schemas.
The name of the Atlas cluster data source to sync. You cannot use sync with a serverless instance.
The name of a database in the synced cluster where App Services stores data in Development Mode. App Services automatically generates a schema for each synced type and maps each object type to a collection within the database.

The current state of the sync protocol for the application.

Valid Options:

  • "enabled"

  • "disabled"

The number of days that the backend compaction process waits before aggressively pruning metadata that some clients require to synchronize from an old version of a realm.
If false, Recovery Mode is enabled for the application. While enabled, Realm SDKs that support this feature attempt to recover unsynced changes upon performing a client reset. Recovery mode is enabled by default.
A list of field names to use as global queryable fields.

A list of field names to use as the indexed queryable field. While this property is an array, Sync currently supports only one one indexed queryable field. Therefore, this array may contain at most one element.

The indexed queryable field must be present in the schema and be the same eligible field type in every collection you sync. The indexed queryable field name must also appear in queryable_fields_names since this is a global queryable field.

{ [collectionName: string]: string[] }
A map from collection names to a list of collection-level queryable fields for each collection.
The date and time that sync was last paused or disabled, represented by the number of seconds since the Unix epoch (January 1, 1970, 00:00:00 UTC).
An array of the names of collections that are defined as asymmetric with Data Ingest, where clients can write data but not read.


Enable Atlas Device Sync


Pause or Terminate Sync