Thursday, September 19, 2013

OakLeaf Windows Azure Website Upgraded to Standard Tier to Avoid Outages from Memory Use

Adding new posts to my Android MiniPCs and TVBoxes Shared Windows Azure Web Site (WAWS) began suspending availability and redirecting to this Black Screen of Death about 15 minutes after each clock hour:

image 

Note: For details about the start of this problem, see my Oakleaf Windows Azure Website Encounters Memory Usage Limit of 9/17/2013.

The Windows Azure Management Portal’s Usage Overview screen showed 390.53 of 512 MB/hour Memory usage one minute after being reset:

WAWSMemoryReset59Minutes

Running a Shared tier WAWS, which has a memory usage cap of 512 MB per hour, costs about US$10/month. A Small-instance Standard tier WAWS, which includes 1.75 GB of memory, costs $0.10 per hour or about $75.00 per month. Fortunately, I’m running the site under a Microsoft Partner Network Cloud Essentials subscription, which includes $100 per month of Windows Azure services. Otherwise, I would have moved the site to a free service, such as Blogger.

Note: The Windows Azure Cloud Essentials benefit is no longer available to new subscribers.

I attempted to upgrade the site from the Scaling window by selecting the Standard Web site mode and clicking Save:

WAWSScalingPageToStandard 

However, I received the following error message because the site was suspended:

CannotSaveWebSiteComputeMode

After waiting about 40 minutes for the site to become available, saving the upgraded compute mode succeeded.

Affect of Tier Change on Reported Site Availability and Response Time

Redirecting to the BSOD instead of returning a 404 error causes Pingdom to not report site downtime and refrain from warning the site owner about the outage. Thus the change won’t affect comparative downtime data reported by my Uptime Report for My Live Windows Azure Web Site: August 2013 = 99.71% and later monthly posts.

I was surprised to discover an extended outage approximately three hours after upgrading the site. Here’s Pingdom’s downtime report:

image

Another member of the Azure Insiders group reported the same problem in the West US data center. A Microsoft representative reported at 8:40 PM PDT that the problem was solved.

It remains to be seen if moving from Shared to Standard tier reduces the site’s response time.

0 comments: