The last few posts I’ve talked a LOT about configuring diagnostics. Much of that comes not because I love pretty graphs, but because I end up working with customers who are troubleshooting problems with applications running on Windows Azure.
Here is MY list of recommendations to help things run a little smoother:
- Keep your diagnostics account separate from your production account. This will help with performance of both production and diagnostics since they won’t be competing for the same storage account.
- Make sure your storage account is in the same data center as your compute. I know. Just saying.
- Make sure you collect the right set of performance counters AND check that you are actually collecting data. Make sure you are collecting the .Net 4.0 counters for ASP.NET where applicable.
- Use either the .wadcfg or PowerShell scripts I’ve talked about here to configure diagnostics. Hard coding it will overwrite any changes you make when an instance restarts.
- Knowing and understanding your baseline workload is important. You should look at your performance on a regular basis, and over a period of time.
- To troubleshoot, you can enable RDP and perform an upgrade of your service. You can then RDP into specific instances to troubleshoot.
- When you are working with Microsoft’s product support, try not to delete old deployments. Once you do it makes finding a root cause more difficult. You can always VIP swap them out and leave them running while they troubleshoot.
- Having more instances running means more people can look at the problem at the same time.
- Check the status of the service at http://www.microsoft.com/windowsazure/support/status/servicedashboard.aspx
- Invest in a diagnostics data viewing tool, such as Cerebrata, or grab a free trail of ManageAxis by following the rabbit hole from the Cloud Cover Show.