Log Analytics alerts aka Azure Log Search Alerts allows you to create a single alert and trigger alert instances per a column. This is possible by creating metric measurement alert. Unfortunately these types of alerts also have a limitation which might be blocking in certain situation. The classical example for that limitation is to create a single alert that will separate instance for free disk space threshold for each computer and disk. The query representation of that looks like this:
Microsoft recently introduced a new model for purchasing Azure Log Analytics. To use this new model you will basically have to enable it on per subscription bases. In short you can either continue to use the old models or flip a switch on your subscription to use the new model. You cannot use both models for different workspaces in your subscription and you cannot move between the old models and the new ones without flipping that switch. Of course flipping that switch is easy as going into the Azure Portal -> Azure Monitor -> Usage and estimated costs blade but what about if you want to do this programmatically? Apparently there is a way to do it that way and in this blog post I will show you how.
I’ve recently stumbled on forum question asking for chart that list only the top 10 resources or computers. Such chart is of course used a lot in performance metrics where for example you want to get the top 10 computers with CPU usage and list them in time chart.
In Azure Monitor we can create two type of alerts for Log Analytics:
Near real-time metric alerts are scoped to specific performance counter and heartbeat events but with Custom Log Search Alerts you can alert on any log in Log Analytics. With Custom Log Search Alerts the alert logic have two types:
- Number of results
- Metric Measurement
In a typical scenario you will use Number of results for logs and events and metric measurement for performance/metric logs. That wouldn’t be a problem if the way the alerts are fired distinguish quite a lot between those. For example in metric measurement you aggregate/summarize results and you alert based on the value from the aggregation/summarization. On top of that different alert instance is fired on each summarized record. In number of results you do not summarize/aggregate and alerts are fired based on the count of the records. For example on 10 records you will get only one alert instead of 10. If you are like me this is a problem as you want to get separate alert instance for your events just like metric measurement alerts.
In this blog post I will show you how to overcome this problem with workaround from the powerful Log Analytics query language.
As you know both OMS Linux and Windows agent send heartbeat events and they are free of charge. The problem is that the interval of these heartbeat events is different for both operating systems. For Windows it is every 1 minute and for Linux is every 5 minutes. I do not know exactly the reason for this decision but I prefer that all my servers report at the same interval. The beautiful thing with the OMS Linux agent that is extendable and configurable. So this blog post will focus on how you can easily change the heartbeat interval on OMS Linux agent to 1 minute.