Make Throttled Request metric (Cosmos DB) available for configuring a unified alert
When looking to migrate our alerting (deployed in ARM templates) to the new alerting platform the throttled request metric is not available when configuring an alert for Cosmos DB resources, we currently have a classic alert using this metric. I have attempted to use an alert based on 'Total Requests' filtering for response code 429 but this doesn't seem to behave the same as the existing Classic alert (maybe due to the only available setting for 'timeAggregation' being 'Count'). Having logged a support request it was suggested I make a request here.
Marking as completed with the response below.
Using TotalRequestUnits with StatusCode 429 worked for me. I believe 408 was a typo.
Status Code 408 is a timeout, as I understand it. Is this really meant to be 408 rather than 429? I'm not having much luck following the link in the answer.
Satish Kumar Rangavajjula commented
You can create new alert in Azure Monitor based on new metrics. The new metric equivalent to Throttled Requests is TotalRequests with StatusCode 408
More information can be found here:
Also, if you have an existing classic alert, you can migrate to new Alert with Classic Alerts migration tool
Razvan Voinescu commented
I have also the same problem. Let us now when will exist a fix.
Marcos Nils commented
I have a similar issue whether I can't seem to find a way to alert on "constant" throttled requests. The Total Request metric only has "Count" aggregation so it's pretty much binary when setting up an alarm based on that (it either exceeded or not). We're getting alerts when we have spikes due to traffic log which we'd like to address by at least being able to do an "average" count of the throttled requests in the last x minutes.
I'm in the same boat. I rely on this alert to keep my applications responsive while at the same time controlling costs. I hope this wasn't removed with the idea of forcing higher, and often unnecessary, increased costs.
Yash Ghia commented
I am looking at similar thing, didn't find any default metric alert to monitor 429 (Throttled requests)