Incidents to trigger on measure split by agents
Apologies if this is covered but I am opening new topic as I just can't find it if so.
Attempting to create an incident that is based on a measure split by agent.
My intent is for the threshold on that measure to trigger on a per agent basis and, therefore, trigger the incident on a per agent basis.
However, this does not appear to be working for me as it will only fire when the total ( no splitting ) breaches the measure threshold.
What is odd is that when the incident triggers for the total measure and you view the details - it shows each agents counts for that measure listed.
Has anyone found a way to alert on a per agent basis using a single measure? I know it could be done otherwise by creating the same measure for each agent then adding all those measures to the incident, but that does not scale when the number of agents is large.
Thank you
Adding onto Brad, we have a use case with my customer for creating incidents based on dynamically generated measures. We'd need to create 800-1500 unique URL monitors to be able to create an incident for each URL. This is a large amount of time, especially when we could easily generate a few URL monitors that test against all of the sites by modifying XML. If we could generate an incident from an individual dynamic measure, that would cut down a lot of time in configuration and maintenance. Adding onto Brad's post here for visibility and if there's anything we might also be missing to implement this. Thanks!
What you are creating is called a "dynamic measures" since you are creating something general, and dynaTrace will then manage all of those dynamic "splits" for you.
You can set a threshold, but it's for the general measure itself - all of the splits will inherit that same threshold. It's not possible to have individual different thresholds for the dynamic elements.
Rob
thank you rob.
However, a point of clarification.
I am not trying to set a diff threshold for each agent group, but rather apply the same threshold for each.
However, the incident will NOT trigger on the inidividual agent basis but only the total unsplit (no splitting ) value of the measure.
I am trying to find out if there is any way to have the incident trigger as each agent breaches the threshold.
Hi,
I am not sure, but maybe something you could check:
If you go into "Edit Measure" of this specific measure and then in the tab "Details", is the checkbox "Create a measure for each agent" checked?
I think this would have something to do with it...
Best regards,
Dennis
Thanks Dennis - I have tried that and to further show what I am attempting including a screen shot - sorry for ugly black scribbling to remove names.
I also tried to change it over to calculate for every agent group and it also did not solve my issue - it continued to only alert on the entire no split measure instead of for each agent.
Before anyone asks or points it out - yes there is multiple agent groups checked on that screen if you would see the scroll down.
Here showing the incident details when setting the threshold above the no spliting total to get it to fire - as you can see it shows each agent that I selected violate the threshold, however, if I change the threshold to just violate a subset of agents, it does not fire for those agents.
Hi Brad,
Ok, I see. Excuse me that I have to ask for more details:
How is your incident setup? Do you have one measure per agent inserted? If so, what did you set in the "logic" column (should be "OR" then)? When I understood correctly though you just have one measure for all the agents (from your description), so maybe you would need to change the "aggregation" column in the incident setup to "count" instead of "sum"? As you have a count measure for web requests the default is "sum". Please note, I am not too sure about my assumptions, just trying to give some ideas to figure it out. Please let me know if you fixed it and how 
Best regards,
Dennis
Dennis, All good thoughts.
However, I am using a single measure (with underlying splitting enabled on agent ). I realize I coudl create a measure for each agent then add with 'or' in the incident to accomplish this but it's not a scalable solution.
I have tried count and sum also and haven't seen any notable change in behavior.
