GeoEvent Server CPU Alerts

[MeleKoneya1]

We have GeoEvent Server processing CAD data from our Fire Department and some partial enrichment of existing tracking features.

GeoEvent uses a lot of CPU but does not seem to be maxed out.      The way that our alerts in Monitor are set up is we get an email at 80% or above which happens frequently but not prolonged.

Can we set the threshold for CPU to a higher number and get fewer emails?      I really don't need to know about the little spikes above 80%, but bigger ones to 100% or even better prolonged periods at 100%

Thank you,

Mele

[CodyPatterson · Accepted solution]

Hey @MeleKoneya1 

You can indeed change these, going to Monitor > Components > <server> you can see the metrics that are measured. Clicking on CPU Utilized or any other item you'd like, you can adjust the monitoring of that by clicking "Edit alert rules".

You can then set the warning and critical amounts at the bottom:

Edit: Along with this, if you'd like to edit the notification settings, you should be able to find them in Administration > Notifications.

Cody

[StevenBeothy]

Just to add to this topic a major improvement that Esri introduced with ArcGIS Monitor 2023 was the Sample interval setting. 

Samples allow you to filter short-term spikes in metric data values and outlier data. For example, when an alert rule is configured to use the average of three samples and the component's Metrics observer is configured for one-minute intervals, the average of the metric data samples must be outside the configured threshold for three consecutive minutes before an alert is opened. If the average of three consecutive samples doesn't fall outside the configured alert threshold, no alert is opened.

This should limit the amount of false positives you receive with brief CPU spikes and is set to 3 as a default for the CPU Utilization metric. As mentioned by Cody, you could increase the Warning threshold from the default of 75% or even the Samples value, but Critical being at 90% may be a good threshold to keep.