- Più recenti
- Maggior numero di voti
- Maggior numero di commenti
So, I had made an incorrect assumption that the Insight "Group" would lump all of my API endpoints together and start reporting anomolies if ANY of the endpoints started to produce errors or faults. For example my endpoints A through D were running successfully, and endpoint E I reserved for errors and faults. So for hours I would use Postman to invoke endpoints A through D to show "good traffic" then when I wanted Insights to start "seeing" faults and errors, I would invoke endpoint E. Well, this was incorrect thinking on my part and found out that the Insights are actually "smarter" than that. So what I did to test was run my postman tests for endpoints A through D and then I deployed a "failing" API container at the third hour of my postman tests which caused endpoint C to start producing faults. This started producing Insights and notifications I wanted. So for anyone having trouble with this, I hope this helps you out.
Hi,
Thanks for reaching out. Are you able to see Insight under Service details after clicking on the affected node ? [1][2]
I would like to highlight one point that X-Ray Insights uses statistical modeling to predict the expected fault rates of services in your application. It will trigger an Insight if a group’s service map contains faults which significantly exceed our forecast.
Insight may miss an anomalous event if there were faults, but the group in question typically sees that level of faults, so it is not anomalous. Also, in case if there wasn’t enough data to train the model and predict anomaly.
As you enabled Insights recently, I would suggest you to wait for some more time for insights to be available.
Feel free to reach out if you continue to face any issues.
Contenuto pertinente
- AWS UFFICIALEAggiornata 2 anni fa
- AWS UFFICIALEAggiornata 2 anni fa
- Perché non riesco ad accedere o visualizzare i dati di Performance Insights in Amazon RDS per MySQL?AWS UFFICIALEAggiornata 3 anni fa
- AWS UFFICIALEAggiornata un anno fa
In my image above of the service map, you can see how I suddenly injected a large amount of errors and faults. This was a sudden thing, it should have been caught it as an anomaly. My process was that I sent hours of data with no faults, and no errors. Then I suddenly started generating faults and errors. I mean, how else can this be tested?? I read that it takes at least 90 minutes of data to train, but I had been sending over 6 hours of data. If faults are happening and we want to be alarmed of those faults BEFORE the training occurs, what use is Insights?