Skip to main content

Data Stores

A Dragonfly Cloud data store represents a Redis Protocol (RESP) endpoint.

To create a data store, on the Data Stores tab, click +Data Store

The minimum configuration consists of name, cloud provider, cloud region and plan.

Note that the cloud provider and region can not be modified once the data store is created.

The plan specifies the provisioned memory and memory to CPU ratio of the data store.
The Enhanced plan has more CPU for the same amount of provisioned memory and achieves higher throughput compared to the Standard plan.

You can modify the data store plan later with zero down time to scale it up or down.

By default the data store will be configured with a public endpoint, TLS and an auto generated passkey, meaning you can securely connect to it from anywhere over the public internet.
To create a data store with a private endpoint, see Security, Networks and Peering Connections.

By default the data store will consist of a single Dragonfly server, to create a highly available data store see High Availability.

Once the data store is created, clicking the data store row will open a drawer with the data store configuration, including the auto generated passkey and a redis compatible connection string.

Once the data store Status becomes Active you can try access it with e.g. redis-cli -u <connection string> PING

To update the data store configuration click the pencil edit button in the top right of the drawer.
Dragonfly cloud performs data store updates with zero downtime.


Dragonfly Cloud supports public and private endpoints

Public Endpoint

With a public endpoint you can connect to the data store from anywhere over the internet. To protect your data store from unauthorized access public endpoints are configured with passkey and TLS enabled.
TLS has some performance impact so can be disabled but it is highly recommended to leave it enabled.
Passkey is mandatory and can not be disabled for public endpoints

Private Endpoint

Private endpoint provides better security, performance and lower latency as the data transports to and from the data stores over private networks and not via the public internet.
Using a private endpoint also reduces data transfer costs applied by the cloud provider.

In order to create a data store with a private endpoint, you must first create a private network, see Networks for more information. Once you have created a private network, you can select it in the Endpoint dropdown box when creating a data store.

Tip: In order to completely avoid data transfer charges, place your data store in the same availability zone of your application. See High Availability for specifying the data store availability zone.

TLS and passkey are disabled by default for private endpoint datastores, but can be enabled.

Durability and High Availability

Eviction Policy

Eviction policy controls the behavior of the datastore when it maxes out its memory.
No Eviction - Items are never evicted and out of memory errors are returned when the data store is full.
Cache - The data store behaves as cache and evicts items to free space for new ones when the data store is full.

High Availability

By default the data store will consist of a single Dragonfly server, this means that in case of software failures, hardware failures or cloud zone outages data is lost and the data store may be completely unavailable.

To increase availability of your data store you can configure it to be deployed in up to three different zones, one primary zone for the master and up to two replica zones. Dragonfly Cloud automatically detects failures and performs failover to an available replica. Select up to three zones in the High Availability dropdown box. The data store master node will be placed in the primary availability zone selected first.
To avoid data transfer costs incurred by the cloud provider, select the primary availability zone to match the availability zone of your application.


BullMQ - Enable this for running BullMQ workloads, this requires you to apply Redis Cluster curly braces syntax for the queue names as described here.

If that is not possible please contact support.


At any time click the chat icon in the bottom right corner to ask anything, really, anything!