Spike detection for Incidents
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
3 weeks ago
Hi,
Did anyone implemented spike detection for incidents (running near real-time) to provide any abnormal increase? I saw one option as PA and trigger notification on threshold breach but running DC jobs, every 5 mins is not recommended.
Any other alternate solution we can use?
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
2 weeks ago
Hi Buddy,
Yes — this is a common requirement, and you’re right that PA alone isn’t ideal for near real-time spike detection. It’s built more for trending and reporting, not minute-by-minute alerting.
Most teams handle this in one of three ways:
If ITOM/AIOps is available: use anomaly detection (Metric Intelligence). It’s designed to spot abnormal spikes against a baseline rather than just fixed thresholds.
Platform-only approach: create a lightweight scheduled check that runs every few minutes, counts incidents in a rolling window, and triggers an alert if it jumps beyond expected levels.
MetricBase (if in use): push incident counts as a time-series metric and use predictive/anomaly capabilities there.
PA threshold alerts will work, but running data collection every 5 minutes isn’t the intended pattern and can create load.
So if the goal is true “spike detection,” I think the best path is:
AIOps/anomaly detection if licensed
otherwise a simple rolling-window monitor script/flow rather than PA collectors.
@VarunK312799563 - if help u answer; please mark Solution Accepted and Thumbs Up
