How to monitor IOPS for an Azure Storage Account

Question

Having used Azure for some time now, I'm well aware of the default 20,000 IOPS limit of an Azure Storage Account. What I've yet to find however is up to date documentation on how to monitor an account's IOPS in order to determine whether or not it's being throttled. This is important when debugging performance issues for applications, VMs, and ASR replication - to name but three possible uses.

If anyone knows the correct way to keep track of an account's total IOPS and/or whether it's being throttled at any point in time, I'd appreciate it - if there's a simple solution for monitoring this over time, all the better, otherwise if all that exists is an API/PowerShell cmdlet, I guess I'll have to write something to save the data periodically over time.

Answer 1

You can monitor your storage account for throttling using Azure Monitor | Metrics. There are 3 metrics relevant to your question, which are

1. AnonymousThrottlingError
2. SASThrottlingError
3. ThrottlingError

These metrics exist for each of the 4 storage account abstractions (blob, file, table, queue). If you're unsure how your storage account is being used then monitor these metrics for all 4 services. Things like ASR, Backup and VM's are going to be using the blob service.

To configure this, go to the Azure Monitor | Metrics blade in the portal and select the storage account(s) you want to monitor. Then check off the metrics you're interested in. The image blow shows the chart with these 3 metrics configured for the blob service.

You can also configure an alert based on these metrics to alert you when any of these throttling events occur.

As for measuring the IOPS for the storage account, you could monitor the Transactions metric for the storage account. This is not really measuring the IOPS, but it does give you some visibility into the number of transactions (which sort of relates to IOPS) across the storage account. You can configure this from the storage account blade and clicking Metrics in the Monitoring section as shown below.