Azure Site - Minimizing Small Frequent Outages

Question

I have a product SaaS, placed on the azure. I have a standalone .net web application and a simple marketing site that is hosted on azure websites. Both were set to the same instance with automatic zoom.

I am concerned about the small frequent blackouts that occur with my websites. Azure says it is expected to be around 10 minutes a week, and that is in the SLA, but I want to minimize this as little as possible.

So, for my marketing site, I changed the instance account to 2 and deleted autoscaling. From what I read, having two copies should really help with any problems - how can azure tell itself :

Please note that Azure websites already have a load balancer feature when switching to another resource for sites in the data center (also known as the region), regardless of website mode.

Sounds great. BUT ... Currently, it does not matter. I have pingdom monitoring on both the marketing website and the main application, and I had 2 2-minute breaks this morning. Both sites had exactly the same problems and were resolved simultaneously.

Am I expecting too much? All documents indicate that this is a good strategy (with the exception of the traffic routing manager) to mitigate this.

thanks

Answer 1

What zoom mode are you working in? If -Basic or- Standard, did you enable the Always On flag?

http://azure.microsoft.com/en-us/updates/azure-web-sites-adds-always-on/

Answer 2

This question seems to be a bit popular, so I am posting an update - in a year ...

My observations so far:

The outages that I experienced about a year ago basically (but not completely) ceased. I think most of them - things that are changing in Lazur, and knock out the site for a few minutes here and there - lately I have not had anything like these issues. I have changed quite a lot, but since then ...

• I always make sure something else is a little long in line
• I make sure that I use a retry strategy to communicate with external services to make the application more robust.
• I do not run more than one instance if I do not need bandwidth - this usually does not help in terms of resistance to the problem that I originally published.
• Now I run one "large" instance. As I became big, I seemed to have a lot less problems. Perhaps this is partly due to the improved VM startup speed or additional RAM / CPU, which makes it more able to deal with problems, I'm not sure.

Answer 3

One way to get higher availability is to create two identical sites in two different regions and use Azure Traffic Manager to switch to another resource from one site to another if one of them decreases.