Azure Sentinel - Logs delay?

Copper Contributor

Hello,

 

I've connected my windows server with Azure Sentinel via Security events data connector.

A few days ago, the delay of getting logs from windows event manager to Azure sentinel logs was 50 secs, now it is 10 minutes.

 

Any idea on how I can see why it takes so long?

 

Thanks!

3 Replies

@FeintBE 

 

To the Events table?

 

Is this all Computers, or just some?  Are they in the same Azure Region as Log Analytics?

 

Factors that affect latency are discussed here: https://docs.microsoft.com/en-us/azure/azure-monitor/platform/data-ingestion-time

 

This query may help to see if its a particular Computer - assuming its the Events table (if not you will need to edit)

 

Event
| where TimeGenerated > startofday(ago(7day))
| extend E2EIngestionLatency = ingestion_time() - TimeGenerated 
//| extend AgentLatency = _TimeReceived - TimeGenerated 
| summarize 
      ['average E2E IngestionLatency'] = round(avg(todouble(datetime_diff("Second",ingestion_time(),TimeGenerated))/60 ),2)
    , ['minimun E2E IngestionLatency'] = round(min(todouble(datetime_diff("Second",ingestion_time(),TimeGenerated))/60 ),2) 
    , ['maximum E2E IngestionLatency'] = round(max(todouble(datetime_diff("Second",ingestion_time(),TimeGenerated))/60 ),2)
  by Computer , bin(TimeGenerated, 1d)
| order by ['average E2E IngestionLatency'] desc
//| render timechart  

   

@CliveWatson 

 

These are the results of the query

res.PNG

@FeintBE 

 

So your average and minimum look good, but the maximum latency was high yesterday (over 10mins).  I assume that is the same computer (behind the red squiggle)? 

Was the computer online yesterday, you should get ~60 heartbeats per full minute?

Heartbeat 
| where Computer startswith "< insert computer name>" 
| where TimeGenerated > ago(1d)
| summarize count() by bin(TimeGenerated,1h)
| render columnchart 

 

Were other computers affected - that would help to rule out Network issues between your machine and Log Analytics?

Was it the Agent compared to the E2E? 

SecurityEvent
| where TimeGenerated > startofday(ago(7day))
//| extend E2EIngestionLatency = ingestion_time() - TimeGenerated 
//| extend AgentLatency = _TimeReceived - TimeGenerated 
| summarize 
      ['average E2E IngestionLatency'] = round(avg(todouble(datetime_diff("Second",ingestion_time(),TimeGenerated))/60 ),2)
    , ['average Agent Latency']        = round(avg(todouble(datetime_diff("Second",_TimeReceived   ,TimeGenerated))/60 ),2) 
  by Computer , bin(TimeGenerated, 1d)
| order by ['average Agent Latency']  desc