I am getting lot of false alerts for JVM heap utilization due to fluctuations in JVM usage.I have used all aggregation for this.I get more than 100 incidents in a day for a JVM. The evaluation time frame is 5 mins.If we increase the time frame there is a chance to miss genuine alerts. Suppressing the incidents won't serve my purpose.Can anyone suggest a method or procedure to get ride of those alerts in future?.
You could try setting your aggregation to minimum while keeping you upper threshold high. This would check to see if the lowest value for your heap utilization is above your threshold. This would only alert if all the values within the time frame were above your upper threshold.
What are the threshold you have applied and what is the aggregation? Try increasing the thresholds and also check if your application has a stable GC or heap utilization pattern (saw tooth).
Thanks for the reply.It is not possible to increase thresholds,since i don't want to change the standard values. GC is not stable.For instance the heap usage(in %) is 90.12,88.76,91.23,92.54,89.87 for 5 mins.The threshold is 90%.If average is taken it will be >90%.Similarly i tried all aggregation except min,I am waiting for data with Min aggregation.Based on my understanding the Min should work.