Solve Logging as a Cross-Cutting Concern with MediatR

[@MilanJovanovicTech]

Want to master Clean Architecture? Go here: bit.ly/3PupkOJ Want to unlock Modular Monoliths? Go here: bit.ly/3SXlzSt

[@3m3rs0n]

It's perfect how you go right to the point and make it easy to understand. Good job Milan!

[@MilanJovanovicTech]

Great, good to know that 😁

[@sergiodurval715]

Your video helped me a lot I'm studying cqrs with mediator and I intend to use this example log in my project Thanks for sharing knowledge

[@MilanJovanovicTech]

Glad it was helpful!

[@ThrottleScotty]

Excellent video, as usual. I really like the format, succinctness and detail. Keep up the good work Milan!

[@MilanJovanovicTech]

Thank you, it's going to become even better this year

[@fernandocalmet]

Excellent topic, I was just thinking in the last days how to implement logging with Mediatr. Thank you very much Milan

[@MilanJovanovicTech]

There you go! 😁

[@nove1398]

Great video. Similar approach to making exception handling.

[@MilanJovanovicTech]

Right! Middleware solves many problems

[@vu5700]

Amazing video sir, i really enjoy your videos , very concise and and concrete, keep it up!

[@MilanJovanovicTech]

Thank you very much!

[@Adronius]

I really enjoy your videos! Short, to the point, with great value. It would be great to see the repository. I'd like to learn more on the Result implementation and how the other MediatR classes looks like. Like for example the UpdateProductCommand. Btw, get also some free time, so that you don't burn out. Working on Christmas day is not a good idea ;-)

[@MilanJovanovicTech]

I covered most of these in some of the previous videos - UpdateProductCommand -> Minimal API Organization - Result -> Validation from Aug/Sep - Repository -> Repository Pattern video

[@kwwx345]

It would be nice to see you make a video on other ways to manage cross cutting concerns like aspect oriented programming with postsharp(or new metalama) or other similar tools.

[@MilanJovanovicTech]

Very interesting, I'll have to add this to my content list! 😁

[@thomaswoods1365]

Really nice Milan. Thank you!

[@MilanJovanovicTech]

Most welcome

[@richardhaughton9633]

Great video again! Could you please make one on logging with ELK stack (Elastic search, LogStash and Kibana) ?

[@MilanJovanovicTech]

Absolutely! I want to cover all the hot topics, but I want to make the videos easy to consume (short) so I can't tackle everything in one video.

[@oyedeoluwafunbi9635]

@@MilanJovanovicTech I am totally looking forward to this!!

[@bugadada]

Much respect for what you do, keep going!

[@MilanJovanovicTech]

I appreciate that!

[@MarcusKaseder]

Why do you add an @ in front of your placeholders? One thing to mention for logging in general: Even if the logs are disabled by config, the cost is still there. If we assume that DateTime.UtcNow will take 1 second to execute, it will also take one second when the log level is disabled. There are some workarounds to avoid this. One is a delegate that only executes if the log level is enabled. Another one is to nest your log into a if that checks the log level and only executes if log level is enabled. Wanted to mention that because some traces can be cost intensive. That traces are mostly not required in production mode.

[@MilanJovanovicTech]

Just a convention I saw somewhere with "@", but it seems it's not even needed. Great point about the cost of non-logged logs, I'll have to bring this up in a future video!

[@MarcusKaseder]

@@MilanJovanovicTech Ah, okay. Wondered if I missed something. It's always about personal preferences 😉 Since those placeholders are reflected by structured logging in JSON or logging targets (logging DBs), it personally feels strange to me to query for @ItemId = 1 instead of ItemId = 1. Great Video as always 🙂

[@MilanJovanovicTech]

@@MarcusKaseder From my testing, they are excluded in logs. At least with Serilog

[@dariogriffo]

Marcus, the @ is identified in "THE" best structured logging library: Serilog as the way to render an object as JSON in your log properties. My suggestion is DONT use Microsoft logging, but use Serilog directly. Microsoft is an abstraction an misses a lot of what you can do with the vendor specific.

[@MarcusKaseder]

@@MilanJovanovicTech @Darío Found the following: The @ symbol before the segment argument is a part of the Message Template specification which means that Serilog must expand the argument's structure as an object. Without this symbol Serilog would just log the result of segment.ToString(). Segment is a class I that case.

[@zikkrype]

Nice one. For me the biggest problem is always splitting LogEvent ID between different application layers (e.g: new EventId(100, "Unhandled"), new EventId(1001, "Payments"))

[@MilanJovanovicTech]

I think that can be handled very nicely by Serilog's LogContext

[@zikkrype]

@@MilanJovanovicTech that's not what i'm talking about. later you filter your logs in kibana or application insights using event(id, name). How to avoid collisions of ids etc

[@syedemad1070]

In vsa will conventional middleware for logging work with mediatr?

[@MilanJovanovicTech]

Yes

[@shirinmonzavi8165]

Hi How to extract information about what handler is handled the request?

[@MilanJovanovicTech]

I'm guessing the RequestHandlerDelegate should have that info somewhere

[@ransandu]

Hi Milan, Thanks for this. Small clarification. The implementation you did is only covering all the inbound request/response. How about logging outbound request/response. E.g Assume we calling a 3rd party application. Want to record what we send and receive in our logs. Any thought on this?

[@MilanJovanovicTech]

If you're calling that API via HttpClient, you can create a delegating handle and introduce the logging in there

[@abdulatifrasulov7291]

Good job. But i have one question. Is this behavior effect API performance if this API should serve large number of requests?

[@MilanJovanovicTech]

Logging will incur a performance penalty, but not knowing what your app is doing is a bigger problem.

[@ederjuniordesouza1618]

Hello Milan great video, thank you for sharing :) I was expecting to see something about MediatR, is it coming in a next video or did I miss something in this video ?

[@MilanJovanovicTech]

Check out this maybe: kzbin.info/www/bejne/rJXMXqNviNmgfZI

[@MilanJovanovicTech]

Also, validation with MediatR: kzbin.info/www/bejne/bmbHqaqaba2te80

[@volodymyrliashenko1024]

I was expecting to see structure logs as a result and than how to use it and what the benefits of having Structured Logs vs plain logs.

[@MilanJovanovicTech]

I'm planning that for the Serilog video which is coming soon

[@volodymyrliashenko1024]

@@MilanJovanovicTech Would be nice if you can recommend application for viewing structured logs locally.

[@fahimehbarzegar5322]

Is it just about log in application layer? How can we use this approach to log infrastructure layer, such as repository? Or what is the best way to log them?

[@MilanJovanovicTech]

You can add a decorator to your repositories, or log within command handlers if you feel you need go

[@BaffourKTonto]

Is Logging the same as having auditing to keep information about which user performed an operation?

[@MilanJovanovicTech]

They serve a different purpose. Auditing is storing data for historical reference. Logs are usually short-lived and are only relevant for a few days.

[@murilomurilo3930]

Where the connection string is stored? I like your videos! ❤

[@MilanJovanovicTech]

Appsettings.json

[@alpsavasdev]

what are the disadvantages of logging in application layer, compared to in API layer (in controller methods)?

[@MilanJovanovicTech]

You can do both

[@ManojRawat-lp4yc]

Is it a good practice to create this pipeline behavior handling class for every .NET core project where we need to log information and errors ? Or is it something specific to MediatR only ? Like if I want to do this for Serilog DB log sinking ? Will it be good design ? Edit ** After going through the Video one more time** I think we can use it anywhere not just with MediatR. But if possible please confirm. Thank You.

[@MilanJovanovicTech]

Pipeline behavior is a MediatR feature. You could implement a middleware if you need this in Web apps.

[@vinland-h1w]

I'm using NLog and I'm confused, It's already in the API project, should I do the same in Application layer? Also, how can this affect the flow if I have validatorsbehavior too. I hope u can respond thanks a lot!

[@MilanJovanovicTech]

I think it's fine to keep it in the Application layer.

[@vinland-h1w]

@@MilanJovanovicTech thank you so much, this helps a lot

[@sauravbhatta5303]

Precise info

[@MilanJovanovicTech]

Thank you!

[@ahsanmehmoodqureshi7035]

please tell the model domain.shared =Result .which fields inside in it

[@MilanJovanovicTech]

gist.github.com/m-jovanovic/aa25b1ae424c985ff8ae696a79b6fe6e

[@winstochurgle9133]

Yo Milan. Thanks for video. I can't enter into Handle method of logging pipeline. Any thoughts? Here's my implementation (i'm using FluentResults): public class LoggingPipelineBehavior : IPipelineBehavior where TRequest : IRequest where TResponse : Result public class CategoryQueryHandler : IRequestHandler, IRequestHandler

[@MilanJovanovicTech]

The problem is: where TResponse : Result Try out with just Result (if it exists) or an interface

[@winstochurgle9133]

@@MilanJovanovicTech Well i tried like this: public class LoggingPipelineBehavior : IPipelineBehavior where TRequest : IRequest where TResponse : IResultBase Because Result inherits IResultBase and it's worked. Thanks!

[@aslanamca8225]

I see cqrs and mediatR even when I want to learn another subject. I hate this situation.

[@MilanJovanovicTech]

If you can't beat them, join them?

[@gerarduab9960]

Thanks. But how to read logs in production environment during certain period of time?

[@MilanJovanovicTech]

Where's your production environment?

[@krccmsitp2884]

In production you would store the logging messages in a centralized logging store, e.g. Seq, Logstash, SolarWind, or a cloud storage system.

[@pbolduc]

@@krccmsitp2884 We use Splunk via Serilog. But I understand what Gerard is asking. For example, a user is reporting an error, but the existing log levels are not verbose enough. It would be useful to be able to turn on more detail logging for a period of time. Previously with trace listeners, you configure your trace listener with a trace switch. You could programmatically at run time change the trace level. You could have an administrative endpoint that you could call to change the level to Verbose (Debug in .NET Core ILogger), and start a timer to reset the level to Information after a short period of time like 5 minutes. I also have use case where I would like to enable logging for one or more users. Although both of these situations do not work well if you are running multiple instances such as on Kubernetes.

[@gerarduab9960]

@@pbolduc that's. Always I listen to add ilogger to logging user data, errors, but who I turn that in order to make it more practice. I think that this data should save it in some kind of log store but during certain periods of time, also this data should be filtered in order to extract metrics. My question it's that, what infrastructures, approach are used in production environments.

[@CyborgT800]

Can you cover Microsoft App insights logging

[@MilanJovanovicTech]

Good idea! I'll add it to my queue

[@10Totti]

Thanks code is on github?

[@MilanJovanovicTech]

www.patreon.com/milanjovanovic

[@rubenguinez92]

I tested this and the content of class LogginPipelineBehavior is not triggered. I really don't know what happened. Of course I've set up DI and the packages installed. (NET 7)

[@MilanJovanovicTech]

Did you setup the pipeline with MediatR?

[@rubenguinez92]

@@MilanJovanovicTech AddMediatr and IPipelineBehavior registered in services/Program.cs, just like you do in the video…

[@rubenguinez92]

@@MilanJovanovicTech In the video, do you use net 7?

[@MilanJovanovicTech]

@@rubenguinez92 Yes

[@MateStier-hd4nv]

I would like to log Result.Value, wow can i implement it?:D

[@MateStier-hd4nv]

I Can with reflection, but not nice PropertyInfo? valueProperty = result.GetType().GetProperty("Value"); object? propValue = valueProperty?.GetValue(result, null); _logger.LogInformation( "Handle Complited {@RequestName}, {@value}", reuqestType.Name, propValue);

[@MateStier-hd4nv]

I would like to make it type safety! Is it possible?:D

[@MilanJovanovicTech]

Why not simply log... Result.Value - what's preventing you?

[@MateStier-hd4nv]

​@@MilanJovanovicTech When add constrain: public class LoggingPipelineBehavior : IPipelineBehavior where TRequest : IRequest where TResponse : Result Mediatr dont use it. I take breakpoint this class , but dont stop -> dont use it. Only use when i use just where TResponse : Result. i want to use Result implementation

[@Maorprog]

Great tutorial. can you please share with us the code in github? Thanks!

[@MilanJovanovicTech]

I only share the code with my Patreon supporters

[@Alexandrplayable]

How to write this if my Result class write how Result??? help meee

[@MilanJovanovicTech]

gist.github.com/m-jovanovic/aa25b1ae424c985ff8ae696a79b6fe6e