Purpose
This article explains how to monitor service health through the platform health-check dashboards and how alerts are generated when services go down.
Overview
The platform includes health-check monitoring through Superset. In this dashboard, users can inspect both:
- a time-series view showing services going up and down over time
- a table showing the current status of services, including logged errors
When one of the services goes down, the platform also sends an email alert so that operators can respond quickly.
When to use health checks and alerts
Use the health-check dashboard when you want to:
- verify whether platform services are currently up
- inspect service availability over time
- identify which services are failing
- review current error information associated with service status
- confirm whether a platform issue is caused by a service outage
Use email alerts when you need proactive notification that a service is down.
How to access the health-check dashboard
To access health checks:
- open the platform landing UI
- click the SuperSet button
This opens the Superset interface containing the health-check dashboard.
What the dashboard shows
The health-check dashboard includes two main views.
Time-series view
This view shows service availability over time. It helps users understand when a service went down, whether the outage was short or persistent, and whether repeated instability is occurring.
Current status table
This table shows the current status of services and includes logged errors associated with unhealthy states. It is especially useful when you want a quick operational summary of the platform at the present moment.
Together, these two views provide both historical and current visibility.
How alerts work
When one of the monitored services goes down, the platform sends an email notification.
This allows operators to become aware of the issue even without actively watching the dashboard. The dashboard is the place to inspect the state in more detail, while the alerting mechanism is what draws attention to the problem in the first place.
Typical workflow for using health checks
A common workflow is:
- receive an email indicating that a service is down, or suspect that a service is unavailable
- open the health-check dashboard in Superset
- inspect the time-series view to understand when the issue started
- inspect the current status table for the affected service
- review the associated logged errors
- continue debugging with logs, Ray, or Airflow if needed
This makes the health-check dashboard a good first stop for operational triage.
Common use cases
Confirming a suspected outage
If users report that a workflow or deployed system is not responding, the health-check dashboard can quickly confirm whether a platform service is down.
Understanding whether an issue is current or historical
The time-series view helps determine whether the issue is ongoing or already resolved.
Identifying the affected service
The current status table helps isolate which service is unhealthy before deeper debugging begins.
Health checks versus logs and metrics
Health checks are different from raw logs and metrics. They provide a higher-level operational summary.
A useful mental model is:
- health checks tell you whether services are up or down
- metrics show behavior and performance over time
- logs explain what happened in detail
Used together, these tools support a much faster debugging workflow.
Best practices
Check health first during suspected incidents
If something appears broken, it is often best to confirm service health before diving into detailed logs.
Use the dashboard after receiving an alert
The email tells you something is wrong. The dashboard helps you understand what and when.
Correlate health state with workflow failures
If Airflow jobs or online requests fail around the same time a service went down, the health-check dashboard can help confirm the connection.
Comments
0 comments
Article is closed for comments.