View and manage alerts
Stack Serverless Observability
For Observability serverless projects, the Editor role or higher is required to perform this task. To learn more, refer to Assign user roles and privileges.
You can track and manage alerts for your applications and SLOs from the Alerts page. You can filter this view by alert status or time period, or search for specific alerts using KQL. Manage your alerts by adding them to cases or viewing them within the respective UIs.
You can centrally manage rules from the Kibana Management UI that provides a set of built-in rule types and connectors for you to use. Click Manage Rules.
To help you get started with your analysis faster, use the KQL bar to create structured queries using Kibana Query Language.
You can use the time filter to define a specific date and time range. By default, this filter is set to search for the last 15 minutes.
You can also filter by alert status using the buttons below the KQL bar. By default, this filter is set to Show all alerts, but you can filter to show only active, recovered or untracked alerts.
There are a few ways to inspect the details for a specific alert.
From the Alerts table, you can click on a specific alert to open the alert detail flyout to view a summary of the alert without leaving the page. There you’ll see the current status of the alert, its duration, and when it was last updated. To help you determine what caused the alert, you can view the expected and actual threshold values, and the rule that produced the alert.
To further inspect the rule:
- From the alert detail flyout, click View rule details.
- From the Alerts table, click the icon and select View rule details.
To view the alert in the app that triggered it:
- From the alert detail flyout, click View in app.
- From the Alerts table, click the icon.
There are four common alert statuses:
active- The conditions for the rule are met. If the rule has actions, Kibana generates notifications based on the actions' notification settings.
flapping- The alert is switching repeatedly between active and recovered states. If the rule has actions that run when the alert status changes states, those actions are suppressed while the alert is flapping.
Alert flapping is turned on by default. You can modify the criteria for changing an alert's status to the flapping state by configuring the Alert flapping detection settings. To do this, navigate to the Alerts page in the main menu, or use the global search field. Next, click Manage Rules, then Settings to open the global rule settings for the space. In the Alert flapping detection section, modify the rules' look back window and threshold for alert status changes. For example, you can specify that the alert must change its status at least 6 times in the last 10 runs for it to become a flapping alert.
recovered-
The conditions for the rule are no longer met. If the rule has recovery actions, Kibana generates notifications based on the actions' notification settings. Recovery actions only run if the rule's conditions aren't met during the current rule execution, but were in the previous one.
An active alert changes to recovered if the conditions for the rule that generated it are no longer met.
A flapping alert changes to recovered when the rule's conditions are unmet for a specific number of consecutive runs. This number is determined by the Alert status change threshold setting, which you can configure under the Alert flapping detection settings.
For example, if the threshold requires an alert to change status at least 6 times in the last 10 runs to be considered flapping, then to recover, the rule's conditions must remain unmet for 6 consecutive runs. If the rule's conditions are met at any point during this recovery period, the count of consecutive unmet runs will reset, requiring the alert to remain unmet for an additional 6 consecutive runs to finally be reported as recovered.
Once a flapping alert is recovered, it cannot be changed to flapping again. Only new alerts with repeated status changes are candidates for the flapping status.
untracked- The rule is disabled, or you’ve marked the alert as untracked. To mark the alert as untracked, go to the Alerts table, click the icon to expand the More actions menu, and click Mark as untracked. When an alert is marked as untracked, actions are no longer generated. You can choose to move active alerts to this state when you disable or delete rules.
Use the toolbar buttons in the upper-left of the alerts table to customize the columns you want displayed:
- Columns: Reorder the columns.
- x fields sorted: Sort the table by one or more columns.
- Fields: Select the fields to display in the table.
For example, click Fields and choose the Maintenance Windows field. If an alert was affected by a maintenance window, its identifier appears in the new column. For more information about their impact on alert notifications, refer to Maintenance windows.
You can also use the toolbar buttons in the upper-right to customize the display options or view the table in full-screen mode.
From the Alerts table, you can add one or more alerts to a case. Click the icon to add the alert to a new or existing case. You can add an unlimited amount of alerts from any rule type.
Each case can have a maximum of 1,000 alerts.
To add an alert to a new case:
- Select Add to new case.
- Enter a case name, add relevant tags, and include a case description.
- Under External incident management system, select a connector. If you’ve previously added one, that connector displays as the default selection. Otherwise, the default setting is
No connector selected. - After you’ve completed all of the required fields, click Create case. A notification message confirms you successfully created the case. To view the case details, click the notification link or go to the Cases page.
To add an alert to an existing case:
- Select Add to existing case.
- Select the case where you will attach the alert. A confirmation message displays.
Stack
Manage the size of alert indices in your space by clearing out alerts that are older or infrequently accessed. You can do this by running an alert cleanup task, which deletes alerts according to the criteria that you define.