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

Metric Events

gunjanmiglani
Visitor

A problem is not detected if we use metric selector but its reported when we use metric key.

The query used for metric selector works fine as it shows correct data in explorer but just doesn't create the problem event.

This is with apigee extension.

15 REPLIES 15

dannemca
DynaMight Guru
DynaMight Guru

Could you provide more details about the configuration settings you are using in this metric event for both?

Site Reliability Engineer @ Kyndryl

 

gunjanmiglani_0-1708689419211.png

With Metric Selector = No Problem raised

Query works as we get data in graph and data explorer correctly.

 

gunjanmiglani_1-1708689546896.png

 With Metric Key = Problem Raised

This is with APigee Extension

 

They both show potential alerts but metric selector one does not create a problem

Are you using fixed thresholds in both? 

Yep.. Its static something like below, we kept everything same for both the scenarios. If the count of errors is more than 1, want to trigger a custom alert. 

gunjanmiglani_0-1708700313448.png

It shows that there is potential alert in both cases but problem is not raised if metric selector is used and problem is raised if metric key is used.

is there any filter used in metric key definition?

Alanata a.s.

Thanks for the reply. No Dimension filter is selected when we use metric key.

ok, but in selector definition you have 4 filters.

Alanata a.s.

Yes.. thats why we want to use selector.

We have 8 apigee apis and by using metric key we get data for all 8 apis which is what we dont want. We want custom alerts for one set of apis which are grouped as request and another set called as response so by using metric selector and filters we want to segregate. 

The issue is query used in metric selector works fine in data explorer and shows potential alert but does not raise a Problem. Hope that makes sense.. 

i see, but there is also "or" filter for 401-404 return codes in selector one

Alanata a.s.

Yes we need that as well to filter the set of apis and trigger an alert if status is one of those!

no issues with advanced model properties? default 3/5/5?
in key definition you seem to sum all 40x return code dimensions and in selector definition you split each 40x code into separate dimension

each of this hints is just guessing, you have to play with settings a bit, b/c it's not always same what seems to be

i suggest you to start playing with some simple selector, that easily can be converted into key definition (f.e. all codes aggregated with sum filtered only per specific apiproxy)

Alanata a.s.

dannemca
DynaMight Guru
DynaMight Guru

The Aggregation you are using in selector is avg, while in key is sum.

Did you tried to change this and observe?

Site Reliability Engineer @ Kyndryl

Yep tried with sum and avg in both cases..  😞 its just the moment we change to metric selector, it just stops detecting problem. 

We are a bit stuck with this. Are there any restrictions on extensions using metric selectors of any sort?? Not sure where we are going wrong. As it works moment we change to metric key and raises a problem. Thanks for the help. Happy to have a call and show if thats helpful sometime today.

g_kat
Helper

Hi,

We are currently facing pretty much the same issue. We have a metric extracted from a log, for which we want to set up a custom event once it passes a specific threshold with an advanced model of 1/3/3(1 violating in 3 minutes windows with 3 dealerting to close). The metric selector is very simple:


log.fallbackbigdata:splitBy():sum:sort(value(sum,descending))


And in the preview we see the data and the potential alert, yet we don't actually get a problem raised.

Featured Posts