Search code examples
azureazure-cloud-servicesazure-diagnostics

What can cause an Azure Cloud Service's Disk Write/Read to spike unexpectedly?

We have a simple worker that picks up messages from a queue and runs a few queries. We don't ever write to the disk ourselves but we do have diagnostics turned on in the roll settings.

Once in a while the the disk write/read spikes and the worker becomes unresponsive. What is the role trying to write to the disk? On the surface it doesn't appear to be a crash dump because those tables and blobs are still empty. Are our diagnostics configured improperly?

Here's an example of a spike we saw recently. It was writing for over an hour! enter image description here

Solution

  • Try enabling remote desktop support in the role configuration in the Azure portal.

    Once the problem resurfaces, log in via RDP and start Resource Monitor. The Disk tab should be able to pinpoint disk IO usage by process and by file.