What's new from Grafana Labs

Grafana alerting has added single-node alert evaluation for high availability clusters. When enabled, only one Grafana instance in the HA cluster evaluates alert rules, reducing query load on data sources.

Normally, when you run multiple instances of Grafana, all alert rules are evaluated on all instances by default. With single-node evaluation enabled, the Grafana cluster automatically chooses a primary instance that is responsible for evaluating all alert rules. Other instances skip evaluation entirely.

With single-node evaluation, the primary instance broadcasts fired alerts to all other instances through the cluster communication channel. This ensures that every instance’s embedded Alertmanager has the current alerts, which is necessary for failure recovery and for the Alertmanager API to return correct data on all instances.

The config.apps and config.panels properties from the @grafana/runtime package are deprecated and scheduled for removal in Grafana 13.2.0, scheduled for the second half of 2026.

Add custom tabs in Frontend Observability that link directly to any Grafana dashboard, so your team can jump to the dashboards without losing context.

Whether it’s a custom performance dashboard, a business metrics view, or a team-specific RUM report, custom tabs let you surface the right context exactly where you need it. Tabs are configured from the Settings page and appear alongside existing navigation tabs once added, making it easy for every team member to access shared dashboards consistently.

We’ve introduced a quality of life improvement when configuring Adaptive Metrics exemptions; now, you create prefix and suffix exemptions, to more easily exempt classes of metrics. Support is available for this in the UI, Terraform provider, and API.

The restore dashboards feature is now generally available! We would love your feedback, you can share it here.

You can restore dashboards you’ve deleted without the need for a support ticket. With the new Recently deleted view, accidentally removing a dashboard no longer means lost work or rebuilding panels from scratch.

Prometheus and Loki data source-managed alerts are being deprecated in Grafana Cloud and can no longer be created in new stacks. Although the pre-provisioned data sources are being depreciated, users can still add their own data sources and can create data source-managed alerts in those data sources. New Grafana stacks now use Grafana-managed alerting (GMA) by default. Datasource-managed alerting (DMA) is not provisioned in new stacks. Existing stacks are not affected.

If your dashboard includes both an API gateway and a database, changing something like an $instance variable affects all panels at once. To avoid this, you might have to split up services across separate dashboards, which defeats the goal of having a unified view.

Dynamic dashboards are now on by default for all Grafana Cloud users and self-managed Grafana instances. Every new and existing dashboard uses the new layout engine, editing experience, and dashboard structure automatically. Your existing dashboards are migrated to the new schema when you open them. No manual steps required.

We’re excited to announce the public preview of Secrets Management for Grafana Cloud k6, a new feature that lets you securely store and use sensitive values in your performance tests. If your tests rely on API tokens, credentials, or any other confidential data, you no longer need to hardcode them into your scripts or pass them around manually.

You can now customize suggested dashboards directly with Grafana Assistant, making it easier to turn recommendations into dashboards tailored to your environment.

When you build a dashboard from a data source, you’ll see a Customize with Assistant option when you hover your cursor over a suggestion. This launches Grafana Assistant with a preconfigured prompt that guides you through adapting the selected dashboard:

As Grafana evolves into a multi-tenant architecture, keeping the operational quality high requires predictable performances. For this reasons we are introducing limits to the number of dashboards and folders that can be created in each Grafana stack. Specifically, we are introducing:

The grafana.com/provenance annotation on alerting notification resources is now correctly read and enforced when writing using the Kubernetes-style API. Previously, provenance was hardcoded to none on all Kubernetes API writes, so the annotation was silently ignored. It is now respected. Setting it requires one of the following permissions:

In Grafana v12.0, we deprecated several Alertmanager configuration API endpoints that rely on legacy single-tenant Alertmanager configuration semantics. In Grafana v13.0 we are removing or restricting access to them:

Remote configuration, a key feature of Grafana Fleet Management, relies on collectors to poll the Fleet Management server at regular intervals for configuration changes.

The problem with this automated approach is that it can be hard to tell when a collector has loaded the new remote configuration or if it experienced a configuration error.