cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 

is dynatrace process crashed detection more nuanced than appmon?

rick_boyd
Mentor

Hi,

I have to alert "informationally" on "Application Process Unavailable (unexpected)" incident from AppMon for things like kubernetes rollouts and even occasional service restarts that for whatever reason don't get picked up gracefully. Is this different for the outside-in approach of the one agent, which goes so far as to actually classify the event as a crash? Can you tell me the difference in profile for these products as far as detection goes?


Rick B

1 REPLY 1

JamesKitson
Dynatrace Leader
Dynatrace Leader

Hey Rick

Someone may be able to provide the info better but here's what I have picked up. It may be a bit difficult to summarize well.

Process shutdowns, crashes, etc... will all be seen separately as distinct events which are going to be fed into the AI problem detection and it will try to find a causation between such events and any impact on services or other problems. So by default you will only get alerts for these if there is an impact on the application itself otherwise you would need to drill in to see these events. If it does see a causation it will report that as the root cause of said problem.

Apart from the above in relation to default anomaly detection there are some options for custom alerting around processes. If you look into custom alerts you can define alerts for given thresholds of number of processes or process groups down to the host level (i.e. if > 2 instances alert) and this can also be done at the process group level settings where you can say under what conditions a problem should be created (should be pretty clear):

Based on what you want I think the middle option above would make the most sense.

Additionally, I would recommend you do some research on process groups as well as it can be a tricky thing to get your head around. These are essentially the groups of process instances over time that metrics are gathered for and are the level that you make configurations (similar to AppMon agent groups). The process group detection in particular may be necessary to get the behavior you want and there is an excellent performance clinic on this:

https://www.youtube.com/watch?v=al6Yh358dFM

Hopefully that helps - let me know if you have questions on particular areas or if anything is unclear.

James