ETL Alert based on lack of activity

[ZachCalvert]

I'm trying to figure out a way to trigger based on the lack of activity.  Every morning we have an ETL job that delivers files, and a client which consumes those files.  This spikes the disk read metric.  If this was in prometheus, I could craft a trend max(disk read) / last 24 hours, and if this maximum rate fell under some threshhold, fire an alert. 

Is it possible with Dynatrace to query max over some window of time, and fire an alert if it falls below that threshhold?

[AntonPineiro]

Hi,

I would say checking metric events.

Best regards

[ZachCalvert]

Hi Anton,

My issue with the metric events is that I don't see a way to trend a maximum over a static duration in time.  I can get the ":max" value, but that is over the set of maximum values for a given point in time with the window being out of my control.  Ideally I could say "trend the maximum identified disk IO usage over the last 24 hours" and that line looks very static until there is a 24 hour window with no activity.  This is what I'm intending to trend.  If I could graph a maximum, of a set window, this becomes straightforward.

Is there a way to trend maximum over a static window of time?

[AntonPineiro]

Hi,

Ummm, I am not sure it that is possible. Maybe some transformation.

It looks like a product idea to me.

Best regards

[ricuevas · Accepted solution]

You can to use metric Process I/O read bytes his metric provides statistics about the I/O read operations of a process. The metric value is a sum of I/O bytes read from the storage layer by all process workers per second. High values help to identify bottlenecks reducing process performance caused by the slow read speed of the storage device.

To create a alert with this metric and disable alert with a maintenance window when the metric timeframe is "normal".   

The metric of the disk not recommend because this metric is affected for all the system and the process. Only you need identify the process that send the files to client and apply what I indicate.