New alert type: Metric Alerts

Hi everyone,

We’ve been developing a new type of alert we’d like your feedback on. Today you can create alerts that are applied to individual Issues – when Sentry receives an event, it checks if alert conditions are satisfied for the Issue containing the event, and notifies you if the conditions match. A couple limitations of this are:

  1. It’s difficult to alert on overall error patterns in a project because conditions are applied to each Issue separately.
  2. It’s difficult to control noise because an alert is generated per Issue that matches the conditions.

Metric alerts aim to alleviate these problems by allowing you to aggregate and set thresholds on all events in your project, regardless of Issue. You can:

  • Filter on any event tags or attributes
  • Set separate warning and critical thresholds
  • Alert when event volume goes below a threshold
  • Send notifications to email and slack (more integrations coming soon)
  • View active and resolved alerts in the app

With Sentry’s performance features, metric alerts will also support metrics like latency and throughput soon.

Here’s what it looks like:

If you’re on SaaS, you can email to have this enabled for your org.
If you’re running Sentry yourself, this feature isn’t available just yet. Docs are available here.

It would be great to know:

  1. Is this going to be useful to you? If so, in what ways?
  2. What additional features would you like?
  3. Anything else
1 Like

Thanks, @adhiraj

We are running an on-premise sentry(version and want to try this feature. I am not able to figure out how to set this “organizations:incidents” flag? Can you help on how to enable this?

@ChillarAnand metric alerts aren’t available for self-hosted Sentry just yet, but we’ll be adding it soon.

Thanks for the clarification, @adhiraj

I was under the impression that there won’t be any difference between cloud service(SAAS version) and self-hosted sentry(latest master of on-premise).

Just wondering if this happens with all new features or just for metric alerts?

I also have an account on the Sentry cloud service. Is there a way to check the version that is running on SAAS?

There’s almost never a difference. For metric alerts there were a couple special things needed to make it work for on-prem, so we’re sequencing the rollout.

1 Like

Metric alerts was a special case as it needs some more Kafka consumers to be set up. We also want to ensure stability for self-hosted Sentry versions as the upgrade cycle for them is much slower (and sometimes people never upgrade them once setup).

You can assume to be running on latest master of getsentry/sentry or something very close to that (barring weekends as we don’t usually deploy on weekends).

I don’t expect metric alerts to be available on the upcoming 20.10.0 release as we only have 2 days to make it happen but it should be out by 20.11.0.

1 Like

The Metric Alerts are limited to “event.type:error”.

We use standard JavaScript setup (JS SDK 5.22.3) and Sentry.captureException() sends “event.type:default”

Two questions:

  1. Why are Metric Alerts limited to error events only?
  2. How can we change the logging to send all events as errors so that we can leverage Metric Alerts?

Metric alerts are now available on self-hosted and also support event.type:default

1 Like