« Playground Elasticsearch billing dimensions »
Elastic Docs ›Serverless ›Elasticsearch Serverless

Differences from other Elasticsearch offerings

edit

Differences from other Elasticsearch offerings

edit

Elasticsearch Serverless handles all the infrastructure management for you, providing a fully managed Elasticsearch service.

If you’ve used Elasticsearch before, you’ll notice some differences in how you work with the service on Elastic Cloud Serverless, because a number of APIs and settings are not required for serverless projects.

This guide helps you understand what’s different, what’s available, and how to work effectively when running Elasticsearch on Elastic Cloud Serverless.

Fully managed infrastructure

edit

Elasticsearch Serverless manages all infrastructure automatically, including:

  • Cluster scaling and optimization
  • Node management and allocation
  • Shard distribution and replication
  • Resource utilization and monitoring

This fully managed approach means many traditional Elasticsearch infrastructure APIs and settings are not available to end users, as detailed in the following sections.

Index size guidelines

edit

To ensure optimal performance, follow these recommendations for sizing individual indices on Elasticsearch Serverless:

Use case Maximum index size Project configuration

Vector search

150GB

Vector optimized

General search (non data-stream)

300GB

General purpose

Other uses (non data-stream)

600GB

General purpose

For large datasets that exceed the recommended maximum size for a single index, consider splitting your data across smaller indices and using an alias to search them collectively.

These recommendations do not apply to indices using better binary quantization (BBQ). Refer to vector quantization in the core Elasticsearch docs for more information.

API availability

edit

Because Elasticsearch Serverless manages infrastructure automatically, certain APIs are not available, while others remain fully accessible.

Refer to the Elasticsearch Serverless API reference for a complete list of available APIs.

The following categories of operations are unavailable:

Infrastructure operations
  • All _nodes/* operations
  • All _cluster/* operations
  • Most _cat/* operations, except for index-related operations such as /_cat/indices and /_cat/aliases
Storage and backup
  • All _snapshot/* operations
  • Repository management operations
Index management
  • indices/close operations
  • indices/open operations
  • Recovery and stats operations
  • Force merge operations

When attempting to use an unavailable API, you’ll receive a clear error message:

{
 "error": {
   "root_cause": [
     {
       "type": "api_not_available_exception",
       "reason": "Request for uri [/<API_ENDPOINT>] with method [<METHOD>] exists but is not available when running in serverless mode"
     }
   ],
   "status": 410
 }
}

Settings availability

edit

In Elasticsearch Serverless, you can only configure index-level settings. Cluster-level settings and node-level settings are not required by end users and the elasticsearch.yml file is fully managed by Elastic.

Available settings

Index-level settings: Settings that control how Elasticsearch documents are processed, stored, and searched are available to end users. These include:

  • Analysis configuration
  • Mapping parameters
  • Search/query settings
  • Indexing settings such as refresh_interval
Managed settings

Infrastructure-related settings: Settings that affect cluster resources or data distribution are not available to end users. These include:

  • Node configurations
  • Cluster topology
  • Shard allocation
  • Resource management

Feature availability

edit

Some features that are available in Elastic Cloud Hosted and self-managed offerings are not available in Elasticsearch Serverless. These features have either been replaced by a new feature, are planned to be released in future, or are not applicable in the new serverless architecture.

Replaced features
edit

These features have been replaced by a new feature and are therefore not available on Elasticsearch Serverless:

  • Index lifecycle management (ILM) is not available, in favor of data stream lifecycle.

    In an Elastic Cloud Hosted or self-managed environment, ILM lets you automatically transition indices through data tiers according to your performance needs and retention requirements. This allows you to balance hardware costs with performance. Elasticsearch Serverless eliminates this complexity by optimizing your cluster performance for you.

    Data stream lifecycle is an optimized lifecycle tool that lets you focus on the most common lifecycle management needs, without unnecessary hardware-centric concepts like data tiers.

  • Watcher is not available, in favor of Alerts.

    Kibana Alerts allows rich integrations across use cases like APM, metrics, security, and uptime. Prepackaged rule types simplify setup and hide the details of complex, domain-specific detections, while providing a consistent interface across Kibana.

Planned features
edit

The following features are planned for future support in all Elastic Cloud Serverless projects:

  • Reindexing from remote clusters
  • Cross-project search and replication
  • Snapshot and restore
  • Migrations from non-serverless deployments
  • Audit logging
  • Authentication realms (native/SAML/OIDC/Kerberos/JWT)
  • Clone index API
  • Traffic filtering and VPCs

The following Elasticsearch Serverless project-specific features are planned for future support:

Unplanned features
edit

The following features are not available in Elasticsearch Serverless and are not planned for future support:

« Playground Elasticsearch billing dimensions »