We have an ASP.NET Core 3.1 web application that is running inside a linux machine and the application is load balanced via GCP Load balancer. Occasionally the dotnet process reaches to 100% CPU and stays there until the application service is restarted. There is no change in the load during this period.
After analysing the thread dump (.nettrace) taken from the linux machine, I see that there are a lot of
System.Net.Sockets.SocketException
exceptions with "Operation canceled" message. There are no further details on where this exception has originated from.
Will this be causing the CPU spikes?
Trace analysis also shows that all the redis commands from inside the application were timed out but the redis server metrics have no errors and all the other applications using the same Redis were working fine. Not sure if this is the result of resource exhaustion due to 100% CPU.