I would like to know is where are anomaly detection Information is stored in ITSI?, I mean any specific Index bucket? or is that a Black Box for us?.I know it is going into "Episode review" but that will not help me . I need to pro grammatically get this information just like I get KPI score from ITSI-Summary. As I need to stop my system alerting when it detects Anomaly.
Secondly, where can I find detailed information about how it is detecting Anomaly?As, I was wondering is there an option to change any setting?
Are you using cohesive AD or trending AD? This is stored in the kv-store. You can configure this in the
mad.conf settings within the SA-ITSI-MetricAD app.
What exactly are you looking for?
Thanks for your time.
Im using Trending AD. what I need is . If you use the feature it shows RED points on graph to indicate its a anomaly. How can I get complete information using Splunk Query ? I cant manually get to any file to review the information.
for example I have a KPI and I want to know whats its Health score. then I can use index=ITSI_Summary and kpi name to get the score value . by running a query, I need similar setup.
Like I said I need to avoid certain steps in a process when model detects anomaly.
This is a limitation of the product. As far as I know, it does NOT write the anomalous behavior to the itsi_summary index. I've been a big advocate for doing adaptive thresholding on a per entity basis which WOULD write to the itsi summary index as it does with the aggregate values. I've also built my own in-house solution of this which works on thousands of entities per KPI. It's much faster than the current AT with a lighter footprint. So my suggestion is to wait for it to become available and keep asking the product team for it
- KPI object in the service object
- A collection in SA-ITSI-MetricAD
- Savedsearches.conf in SA-ITSI-MetricAD
Computational Middle Work:
- there is an index for it called anomaly or something defined in SA-ITSI-MetricAD
Final resultant Anomaly:
- it's a notable event like any other, so tracked alerts index and then the episodes index