The failing workflows dashboard helps you proactively identify workflows that may be experiencing reliability issues. By highlighting patterns rather than isolated errors, this feature allows you to focus your investigation and remediation efforts where they are needed most.
What makes a workflow "failing"
A workflow is flagged as failing if it meets the following criteria within the last 7 days:
Active status: The workflow is not paused or deleted.
Main branch activity: It must have published commits and workflow runs.
Recent errors: The latest completed run resulted in an error.
Minimum activity: The workflow has had at least 3 runs.
Failure thresholds: The workflow meets one of the following conditions:
Streak: At least 2 consecutive failures.
Rate: At least 3 failures with a failure rate of 50% or higher.
How to use the dashboard
The dashboard provides a quick summary and detailed insights into your workflow health.
Summary card
The summary card displays the total number of failing workflows that need your attention. A red sparkline shows the trend of failed runs over the last 30 days, giving you immediate visibility into recent reliability patterns.
Detail modal
Click the summary card to open a modal with a detailed view:
Workflow list: See a table of all workflows currently flagged as failing.
Key metrics:
Failing for: Shows how long the workflow has been experiencing issues (in days, hours, or minutes).
Failed (7d): Shows the ratio of failed runs versus total runs in the last week.
Sorting: You can sort by the longest failing duration or by the volume of failures to prioritize your work.
Each row in the table is interactive, allowing you to click directly into a specific workflow for further troubleshooting.
Technical details
Time window: The system monitors runs from the last 7 days.
Data sources: The dashboard aggregates information from workflow metadata, daily run statistics, and individual run status monitoring.
Scope: Only production runs on the main branch are counted; test or one-off (ephemeral) runs are excluded to ensure the data reflects your live workflow performance.