In your Azure Cognitive Search solution, your users reported errors and concerns such as “error code 503” or “Too many requests. Please try again later”. You have already enabled the diagnostic logging in Azure Portal to log the events and metrics. Analyzing the error, you observe that queries were dropped while a replica was taken out of rotation during indexing. Which is the most appropriate query metric would you check for such a scenario where queries are dropped? (select one answer choice)

Answer options:

A.Queries Per Second
B.Throttling
C.Search Latency
D.Azure Diagnostic

Correct Answer: B
Option A is incorrect because Queries per second (QPS) is the metric used to measure the queries executed in one minute. You can get values such as sum, average, cound, max and min values through this metric in Azure Portal.
Option B is correct because the Throttling search queries metric is used to identify the percentage of queries dropped in a given interval. Throttling is likely during the events where resources are consumed to an extent that they are unable to process additional queries and hence drop the queries. The metric also provides the values such as average, max, min, count and total.
Option C is incorrect because Search latency is aimed at identifying the time taken to complete a query and figure out the latency involved in the processing. It also gives values such as average, min and max latency.
Option D is incorrect as the Azure Diagnostic extension is used for monitoring the metrics of a guest operating system and storing them in Azure Storage. These metrics are used by Azure Monitor.
Reference: 
To learn more about monitoring search queries in Azure, use the link given below: 

https://docs.microsoft.com/en-us/azure/search/search-monitor-queries