Empty cells in the training file are used to create generic matching rules. They represent a wildcard-like behavior, allowing you to apply an action to all values of that attribute when no specific value is provided.
This is especially useful when you want to define fallback behavior or catch-all logic for alert conditions that do not require strict filtering.
Purpose of Empty Cells
When a cell is left blank:
- The machine learning engine interprets it as “match all” for that column.
- The rule becomes more generalized, applying to all alerts that meet the other criteria.
This approach helps reduce file size and maintenance by eliminating the need to list every possible value for attributes that don’t affect the outcome.
Matching Behavior
When a new alert is triggered, OpsRamp’s alert engine:
- Evaluates all rows in the training file.
- Prioritizes rows with the most exact matches.
- If multiple rows have the same number of exact matches, it selects the first matching row.
This means that specific rows (with fewer empty cells) take precedence over general rows (with more empty cells).
Best Practice
- Always test the training file with a limited scope before applying it broadly to avoid misrouted or suppressed alerts.
- Order your training file from most specific to most generic. This ensures that precise rules are applied first, and broader rules act as a fallback.
For a practical example demonstrating how empty cells are used to define fallback rules, see the example scenario in Creating a Training File