How do we tie an expensive SQL Server query back to the part of C# / EF Core code that generated it?
I'm looking for a reliable strategy that will allow me to relate a performance problem that the Azure portal has highlighted, back to some query in the C# .NET 7 / EF Core 7 app that has generated that query.
In Azure I see that some query is routinely in the top expensive/slow queries
SELECT * FROM SomeMassiveTable
So, there's no WHERE
clause - I could start hunting through the codebase for all locations where db.SomeMassiveTable
is used and chase them out to see if anywhere is enumerating it without applying a .Where
but I'm curious to know if anyone out there has developed a better approach.
Is there a way of including an intent kind of comment with a query in a way that would show up in a trace?
For example, perhaps when the developer does something like
db.WithIntent("Cache the whole SomeMassiveTable locally into redis").SomeMassiveTable.ToList()
the Azure portal would complain the top query is:
--Cache the whole SomeMassiveTable locally into redis
SELECT * FROM SomeMassiveTable
so we could search that comment and find it in the codebase.
How about a way of grabbing a generated query (interceptors) and logging the stack trace if the query has no WHERE keyword present? Or getting some feedback from the (local dev) SQL instance itself that it's just run an expensive query, "and here was the stack trace of the C# side that led to it..."?
I can think of various strategies to try, but I'm trying to find out if this is a problem someone has found a good solution for already
In summary, is there a good, easy to implement, built in approach for tracing problematic queries back to the originating code?
Comments
Post a Comment