Last night we had 45,000 failed attempts to access an SQK Server database on a high profile site, which seems to correlate with portions of the application ceasing to work last night. However, when we pushed the main 3 DLLs live this morning, the issue was immediately corrected.
Why would failed attempts to login to the database cause an issue which would be so easily corrected by re-uploading DLLs, and what can be done to prevent this problem from occurring?
Uploading the DLLs will cause the application to reboot, hence it starting to work again.
Lots of failure could cause the application to fail through logging, or IIS seeing multiple failures and shutting down the worker process, or a number of reasons.
Failure to correctly dispose or return database connections to the ADO.NET connection pool in the application might be at the origin of this problem. When you upload DLLs to your site (I assume in the bin folder) the application pool restarts and frees all managed and unmanaged resources returning the application to normal. If incorrect connection handling in your application it might not take longer for this to reproduce.