Action recommendation
Site Reliability Engineers are the glue that holds the application together and are essential to keeping the product up and running at all times. It is critical for them to keep the Mean Time to Repair (MTTR) as low as possible.
When an event occurs, the action framework employs the following algorithms and integrations to match actions to events.
Algorithms
Instana uses pre-defined algorithms based on an event signature that is created when an event occurs. There are three types:
Event similarity
When an event occurs, the event signature is matched against event signatures of the events that are configured in policies. A score is calculated and normalized. The action that is configured in the policy is given this event-to-event score.
Natural Language Processing (NLP)
When an event occurs, the event signature is matched against action signatures in the action catalog. A signature includes the name, description, entity type, and tags. A score is calculated and normalized.
To improve an action's score for an event, enhance the action description and add tags to the action that better align with the details of the event.
Success rate
When an event occurs, actions are given a score based on the success rate for the action when run against past occurrences of this event.
Policies
When an event occurs that matches the conditions specified in a policy trigger, the policy is listed in the Recommended actions table with High confidence.
Turbonomic
If the entity referenced by the event has IBM Turbonomic actions that are associated with it, these actions are recommended based on their NLP score. IBM Turbonomic actions are included only if you configured the integration with IBM Turbonomic. For more information, see Integrating with IBM Turbonomic.
Using recommended actions
Action scores are grouped into a confidence category of Low, Medium, or High. The Recommended actions table displays policies associated with this event, in addition to actions with Medium or High confidence. Policies are listed first, followed by the other recommendations, which are sorted by their score.
When an incident has a Probable root cause section with supporting evidence, the Recommended actions section has a Context for drop-down menu to provide context for the recommended actions. By default, Triggering event is selected in the drop-down menu. However, you can change the context to receive recommendations for the probable root cause. The Generate with watsonx AI feature uses the selected context to suggest recommended actions.
You can view and run recommended actions by using the action's Options menu.
To create a policy from a recommended action, open the Options menu and select Create policy.
Intelligent remediation with watsonx
You can optionally use generative AI with watsonx to create new actions for the event. For more information, see Intelligent Remediation: live action generation with watsonx.