You can configure data source permissions to allow or deny certain users the ability to query, edit, or administrate a data source. Each data source’s configuration includes a Permissions tab where you can restrict data source permissions to specific users, service accounts, teams, or roles.
- The `query` permission allows users to query the data source.
- The `edit` permission allows users to query the data source, edit the data source’s configuration and delete the data source.
- The `admin` permission allows users to query and edit the data source, change permissions on the data source and enable or disable query caching for the data source.
By default, data sources in an organization can be queried by any user in that organization. For example, a user with the `Viewer` role can issue any possible query to a data source, not just queries that exist on dashboards to which they have access. Additionally, by default, data sources can be edited by the user who created the data source, as well as users with the `Admin` role.
You can assign data source permissions to users, service accounts, teams, and roles which will allow access to query, edit, or administrate the data source.
When you enable query and resource caching, Grafana temporarily stores the results of data source queries and resource requests. When you or another user submit the same query or resource request again, the results will come back from the cache instead of from the data source.
When using Grafana, a query pertains to a request for data frames to be modified or displayed. A resource relates to any HTTP requests made by a plugin, such as the Amazon Timestream plugin requesting a list of available databases from AWS. For more information on data source queries and resources, please see the developers page on [backend plugins](https://grafana.com/developers/plugin-tools/key-concepts/backend-plugins/).
The caching feature works for **all** backend data sources. You can enable the cache globally in Grafana's [configuration](../../setup-grafana/configure-grafana/enterprise-configuration/#caching), and configure a cache duration (also called Time to Live, or TTL) for each data source individually.
Storing cached queries in-memory can increase Grafana's memory footprint. In production environments, a Redis or Memcached backend is highly recommended.
When a panel queries a data source with cached data, it will either fetch fresh data or use cached data depending on the panel's **interval.** The interval is used to round the query time range to a nearby cached time range, increasing the likelihood of cache hits. Therefore, wider panels and dashboards with shorter time ranges fetch new data more often than narrower panels and dashboards with longer time ranges.
A panel's interval is visible in the [query options](../../panels-visualizations/query-transform-data/). It is calculated as follows: `time range / max data points`. Max data points are calculated based on the width of the panel. For example, a wide panel with `1000 data points` on a dashboard with a time range of `last 7 days` will retrieve fresh data every 10 minutes: `7d / 1000 = 10m`. In this example, cached data for this panel will be served for up to 10 minutes before Grafana needs to query the data source again for new data.
You can configure a panel to retrieve data more often by increasing the **Max data points** setting in the panel's [query options](../../panels-visualizations/query-transform-data/).
Query caching works for Grafana's [built-in data sources](../../datasources/#built-in-core-data-sources), and [backend data source plugins](https://grafana.com/grafana/plugins/?type=datasource) that extend the `DataSourceWithBackend` class in the plugins SDK.
To verify that a data source works with query caching, follow the [instructions below](#enable-and-configure-query-caching) to **Enable and Configure query caching**. If caching is enabled in Grafana but the Caching tab is not visible for the given data source, then query caching is not available for that data source.
Some data sources, such as Elasticsearch, Prometheus, and Loki, cache queries themselves, so Grafana _query_ caching does not significantly improve performance. However, _resource_ caching may help. Refer to [plugin resources](https://grafana.com/developers/plugin-tools/key-concepts/backend-plugins/) for details.
You must be an Org admin or Grafana admin to enable query caching for a data source. For more information on Grafana roles and permissions, refer to [About users and permissions](../roles-and-permissions/).
You can optionally override a data source's configured TTL for individual dashboard panels. This can be useful when you have queries whose results change more or less often than the configured TTL. In the Edit Panel view, select the caching-enabled data source, expand the Query options, and enter your the TTL in milliseconds.
To configure global settings for query caching, refer to the `caching` section of [Configure Grafana Enterprise](../../setup-grafana/configure-grafana/enterprise-configuration/#caching).
To disable query caching for an entire Grafana instance, set the `enabled` flag to `false` in the `caching` section of [Configure Grafana Enterprise](../../setup-grafana/configure-grafana/enterprise-configuration/#caching). You will no longer see the Cache tab on any data sources, and no data source queries will be cached.
If a data source query request contains an `X-Cache-Skip` header, then Grafana skips the caching middleware, and does not search the cache for a response. This can be particularly useful when debugging data source queries using cURL.