AutoScale Azure crash notifications due to lack of metrics available

Question

In the past few days, we began to receive many emails from Azure, saying the following

Theme:

[AUTOSCALE FAILED] - Auto-scaling set for service name due to lack of available metrics

Message body:

Autoscale has not been able to read any monitoring data ... We recommend monitoring the responsiveness and capacity of your service, to make sure that there are currently enough machines deployed to meet your goals. We will attempt to autoscale as soon as the role instances are available 

We have not made any recent changes, so we are not entirely sure what this means or whether we should take any action. Can anyone shed some light on this?

Answer 1

The original question asked a while ago, but I started to receive this message this morning, so I add my experience.

Digging into the azure portal, I was able to make sure that my cloud service is not actually scaled, so at the moment I turned off autoscaling and left my instances installed at my maximum size. Continuing to search for a portal for tips, I noticed a warning at the bottom of the screen that led me to this message from Microsoft:

We are in extended recovery for this incident. Starting from 8/5/14 to approximately 16:50 UTC, Cloud Services customers with auto-scaling services in the Eastern United States may see errors that indicate that their Auto-Scale metrics cannot be read. Engineering teams deploy their proposed mitigation solutions and continue to monitor the presentation system. An update will be provided after 6 hours or as an event warrant.

In short, for me this is not what I did wrong. This is a problem with the auto-scaling service that Microsoft needs to fix.

Hope this is helpful to others.

Answer 2

You can usually find answers to cryptic errors like this on the Azure status page: http://azure.microsoft.com/en-us/status/

In particular, in the "History" section