0

I am trying to set up a new custom condition for Azure monitor alert rule, but when I enter my kql query it doesn't show the expected data. When I run the same query in Logs it outputs 9 rows that fulfil my condition, but for some reason, no data are shown in the Monitor Alerts.

I can see that the problem is in the last condition | where Anomaly has "1" as I get data when I delete this condition - but I need to have it included in the query (or at least a similar version of it). Any suggestions? (I have tried also contains and == but it gives the same problem)

ADFPipelineRun 
| where TimeGenerated >= ago(30d)
| where PipelineName startswith "Dataflow"
| extend customerName = extractjson("$.customerName", Parameters, typeof(string))
| extend customerBranchName = extractjson("$.customerBranchName", Parameters, typeof(string))
| extend databaseName = extractjson("$.databaseName", Parameters, typeof(string))
| join (ADFActivityRun
        | where ActivityType == "Copy" and Status == "Succeeded"
        | extend RowsCopied = extractjson("$.rowsCopied", Output, typeof(int)))
    on CorrelationId
| summarize AggregatedValue=any(RowsCopied) by customerName, customerBranchName, databaseName, PipelineName, bin(TimeGenerated,1d)
| order by TimeGenerated
| summarize EventCount=make_list(AggregatedValue),TimeGenerated=make_list(TimeGenerated) by customerName, customerBranchName, databaseName, PipelineName
| extend (anomalies, score, baseline)=series_decompose_anomalies(EventCount, 5, 0, "avg")
| extend Anomaly = array_slice(anomalies,0,0)
| where Anomaly has "1"

Thanks for any good ideas and help :)

Martina Dolinska
  • 1
  • 1

2 Answers2

0

The most general answer: start by working backwards and validate your assumptions.

remove the final | where... line and see what the query returns. does it have 1s?

has and has_any and contains all have subtly different semantics, so you may need to use one or the other or somethin.

if your result doesn't have 1s, then work back one more line, is your array_slice call return the items you think it does?

if you just want the 0th item, why even use slice? why not just use Anomaly=anomalies[0] ?

without having your exact data set, there's no way for us to reproduce the query /results exactly.

John Gardner
  • 24,225
  • 5
  • 58
  • 76
0

The problem, in the end, was in the Azure Monitor Alert function. The custom alert has a predefined time range over which it evaluates the query and it cannot be manually extended.

In the predefined period given by Azure, there were no records found (it was too short to evaluate whether anomalies occur..)

I solved it by creating a power bi report that uses M query above and I used an alert function offered by the power bi service.

Martina Dolinska
  • 1
  • 1