We would like to implement a custom alert rule to detect high memory usage in a a given process (e.g., a specific NGINX process). I am not totally clear, however, whether the metric value aggregation used in the rule would be based on (split by) each process group instance or for all instances in the process group. I fear that the latter is the case, but would like to have this confirmed. A memory usage alert should of course be related to a specific process instance, but I am not clear if that is the result we would get.
Also, I can anyone tell me why metric names shown in the Dashboard configuration screen are different than those (in some cases) shown in the Alert configuration screen? For example, in the dashboard we see a "Technologies->Generic->Process Memory" metric while in the alert configuration screen we see "Technologies->Generic->Memory->WorkingSetSize".... both show up as "Process Memory" when selected... whether they are the same thing, and why they are named differently if they are the same is not clear to me.
Solved! Go to Solution.
When problem will show up you will see particular process instance affected.
about second question, process memory is whole memory collected by process. For Java for example it is heap and no heap memory together. About this second metric I’m not sure.
OK that is good to know, thanks. The reason I suspected that the threshold would be evaluated against the average of ALL instances, rather than individually, is that the Dynatrace GUI shows a graphic, when you configure a custom alert, in which you can view the metric and whether alerts would be generated for specific thresholds... this graphic shows only a single AVG value despite the fact that the rule is based on a process group name that has 2 instances in it.
In some cases the metric names are different, we are aware of that and will soon switch all UIs to the consistent names shown in the custom chart metric picker or the custom events metric picker.