It looks like really worth diving into and buying his course, even though it is a bit expensive here in Mexico, but it is worth to save for it. For now it looks really difficult to understand the example since it is one quite complex, a very simple one would have been nice.
This is probably the last step before passing into Microservice Architechture. Best to use it if you're right on the edge of creating separate teams to develop the project. A bit too much separation of concerns if you have a small application. For example, projects Domain/Application/Infrastructure could be all in the same project and only separated by folders, messaging between countext is also very strict in this case and adds a lot of overhead, resulting in difficult/painful to change. So I would say, don't start with this if you're not sure of your bounded context and event generated.
This sample project is a bit dogmatic, I have to say. I'm working on a video series with an improved architecture that should be coming out in a few weeks
Having layers in one project would mean zero isolation between them, which could lead to some programmers referencing between layers in a wrong way. Watching over them would be a pain
@@maxpuissant2 people make mistakes. If you can prevent them technically, then you should do it. It's like programming in dynamic instead of static languages
@@Eugene.Golubev I don't believe you make that mistake by error but if you want a safety net you should write a test, it will be more flexible and explicit.
You say that it is used MassTransit with memory transport. But later you show database tables for inbox/outbox and event handling. Is they part of MassTransit ? Does MassTransit always use db tables despite the way of transport messages ?
Hi Just recently started your videos. Great content. A couple of questions 1. Do you always use DDD in all the projects ? Or that depends on specific use case. If yes then what will be those 2. I saw you used UserPermission. ReadUser or UserPermission. ModifyUser, these are string constants as i saw. Are you saving these permission in some kind of token like cookie or jwt. Just a thought wont it make a token very large with all the permissions or if it is saved in Db then it will suffer from performance as it puts more load on database.
1. Context-dependent - but I do tend to use it in my examples 2. I showcased this in a separate series: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-4cFhYUK8wnc.html You can simply query the permissions based on the User ID when the API is called, and store the values in a cache for quick retrieval in the future.
Yes, some updates coming in January. Subscribe to my newsletter to get notified! It won't be just .NET 8 - there will be new chapters and some re-recordings
Thank you for your amazing content. If you have the time, I would be grateful if you built a complex library containing a lot of abstractions, so we can see a good example of modeling. Thank you for your time.
Great video Milan this is exactly what i was looking for as i was doing my own Modular Monolith project and i needed some answers on how to handle some stuff in it, and you did indeed answered me with this video great content and i hope that we can see more about this topic in next weeks
Cool video :) Nice to see this architecture coming back, 10+ years ago when I was doing PHP, there was a phase where frameworks were moving from MVC to so called HMVC, which was a similar concept and idea. Sadly microservices took over, nice to see somebody pushing this again :)
Hello :) create user trigger one event and return the user. Then you make bind tap ... for validation. But if this user is not valid you return errors but the event is allready save ?,
Great content Milan! I've been implementing this architecture myself and got to a very similar solution structure. The main difference is that your "IntegrationEvents" project is my "Contract" project. That includes not only integration events, but also the definitions of commands and queries. This allows me to query specific data from a different module in a clean and decoupled way, without the need of making the local copy using integration events. I use it when the query is not frequent, otherwise I stick to your approach of duplicating the data using events. I don't currently use it for commands (prefer event+subscriber for that) but I don't see why I couldn't. It's the module public API after all. I also tried the railway architecture you are using here but eventually got back to using exceptions and filters. I couldn't see the gain and all the bind,tap,map makes the code less readable, not more.
@@MilanJovanovicTech Right now I'm good with Mediatr, but I hide it behind an ICommandBus and IQueryBus just in case I need something else in the future.
Hi @Milan, Thank you for this kind of video, For modular monolith how do you deal with UI / Pages of a web app that has datas that getting on multiple modules? A sample page is displaying a listing of added cart items but on the same page Im also displaying the user information like name, address, mobile number etc.
@@MilanJovanovicTech thank you for your insight. In typical MVC that will use modular approach, is it correct to do composite ui? for ex. component ui that present in modules(cart and profile) then consume it to host application (Pages/Cart.cshtml)
I understand the concept, but I’m curious about four things: 1. In the database design at 10:11 how would a customer register for a training be stored? Can I store the userId in the trainings module? 2. How to handle a database transaction, where at least two modules must do something before persisting 3. How to split off a module into a micro service 4. How many endpoints on average would be the maximum size of a module? Thank you for the video
1. Yes, you can store the UserId (or Client - as it's referred to in that module) 2. No cross-module transactions. You treat them as separate transactions. 3. Pull the code out, add a reverse proxy, done. 4. No such metric, you decide
Thank you for a such halpfull video. Could you please share the code used in this video and attach the link to it under the video? It would be very useful to examine it.
How are integration event handlers registered? Is there a section of code that registers a particular integration event with one or more handlers that are interested in this event from other modules?
Thanks for your great content! I would be really interested in an example of the modular monolith approach with the vertical slice architecture. Also, would you recommend to start with the vertical slice architecture in a new project and with a team, that has not much experience with modular monoliths yet?
The two aren't mutually exclusive, nor do they require each other. You can definitely start with VSA. Modular monolith is more about solving the interaction and coupling between modules.
If you don't mind me asking Milan, how would you handle exposing such a monolith if one needs an auth server (like aws cognito or identity server from duende), email service like SendGrid, payment service like Stripe ... would you put upfront, before the frontend, a gateway of sorts (just like for microservices) and reroute calls to those services and make them external APIs talking to your database ... or would you create modules for those and also put them inside this monolith ? I never see anyone discussing Gateways and Modular Monoliths, but just because you have a modular monolith does not automatically imply all your apis and services should be inside of it I assume ... ? I'm a bit lost here on what I should do. I expect up to 100 EF queries (GraphQL queries) per second so I cannot say this is such a huge project that I would need microservices, but not too small either. And I like having a gateway that can do rate / ip limiting and also add correlation IDs ... but to me it feels like that App project is a gateway on its own :D Please help answer this pleb xD
But these are all third-party APIs, right? You typically call them using HttpClient or using an SDK if it's provided. Which is why this part sounds confusing: "would you put upfront, before the frontend, a gateway of sorts (just like for microservices) and reroute calls to those services and make them external APIs talking to your database" I'm recording a bunch of Modular monolith videos this month, and I'll make sure to cover gateways P.S. Help me clear up my confusion, and I'm sure I could help
Not sure if I should call them 3rd party, since they are APIs I will build from scratch for my pet project that indeed wrap the 3rd party sdks from the respective cloud offerings like stripe or sendgrid. I just wish I knew how to approach hiding frontend communication to the backend services from users (clients), by only exposing a gateway … when it comes to “modular monoliths”. Or perhaps it would imply a mix of microservices (gateway, auth, email svc) and modular monolith… Loved your Pragmatic Clean Architecture Course btw 🎉 Looking forward to your vids! Cheers!
Hi Milan! This is what I was waiting for. I greatly appreciate this video, which has helped me significantly clarify some doubts. I hope you continue uploading more videos about how you implement the concepts of CA and DDD in different architectures. Greetings!
Hi Milan Some of your subscribers are from India. And i am also from India. In India we are not able to get access on Source Code. As in our country patreon is not working. Any other way to share the source code with us. So we can clearly understand the flow and other things. Humble request pls do something for us.
@milan, you've mentioned communication between modules is done via an inmemory bus provided by masstransit, which is fine for a fire and forget, but can you explain how we'd go about the retrieval of data from a module. example, module A needs data from module B during the execution time within a method in module A. without direct calls, I cannot see how a pub/sub implementation would work in this scenario
"which is fine for a fire and forget" - it's not fine. You will lose all your events in case of reboot or crash. Async communication in a monolith is very hard. You need outbox pattern and idempotency on the input and all other complexity borrowed from microservices.
Modular is a great way to separate domains within the same monolith, and we can constrain direct access to domain models through assembly properties. Also, another positive thing is even though you are running monolith locally you exclude certain modules that are not dependant from your module and thus cut the build time.
Hi Milan, i think i am too late to ask something, but i have one question about how to organize the modules using ddd. So, i have bounded context, let's say catalog, within this bounded context has 3 sub-domains inventoryManagement, catalogManagement and princing. My question is, each of these sub-domains should be a module? If we use clean architecture, each of these sub-domains should have it's own domain, application and infra layers?
@@MilanJovanovicTech Hi, If my bounded context is large and has 10 aggregates, is it a sign i'm doing something wrong or it's normal and i should put all those entities on the domain folder of that bounded context?
Hi @Milan, very nicely put all things together. I have one question, instead of one executable to expose endpoints of all modules, can't we separate the executables as well per modules? There will be multiple API applications in that case, it might distribute the loads as well. Will it be still monolith if not what it calls?
Thanks for this. Clearly, there was a lot of work involved. It is a nice reference that brings in messaging, functional programming, layers in slices, etc. Unfortunately, the entire purpose of Modularization and Bounded Contexts has been completely obviated by the massive duplication of data maintained in the 2 database schemas. Using integration events to couple Modules and Bounded contexts is odd. Common, but odd. What is the reason for creating this coupling?
@@MilanJovanovicTech With all due respect Milan, that makes zero sense. It is precisely the fact that they duplicate and sync data that makes them completely dependent. That's what coupling is. That's what dependence is. Perhaps, I am misunderstanding something you are doing. Why do you need FirstName, LastName, email in both services?
@@vincentcifello4435 I look at it from a perspective of a single module. It has all the data it requires to function. It can continue functioning even if the other module is down/unavailable. Disregard for a moment that the modules are publishing integration events to pass around this data. Having said that, this is an example project where I purposefully pushed some things to the extreme. So look at it from that perspective also.
@@MiningForPies huh? If you change the FirstName field in User from 50 characters to 80 characters, then you have to change the FirstName field in Trainers. That is coupling.
All of this code, albeit well engineered, is really to replace the same kind of functionality which in a database-centric system would be one or two stored procedures. It is a lot of overhead. Each project might have to have physical project files and folders residing at top level within an overall solution folder in the file system and the more projects and modules there are the more subfolders there will be in that solution folder. How scalable is this, do you think?
Stored procedures are unmaintainable. Database centric systems are unmaintainable. Avoid them at all cost. There is a reason it is *far* from the norm nowadays.
I get where you're coming from, but this is more about logical separation and organization within a monolith. While keeping track of inter-module dependencies and database dependencies. It's actually a very "pretty" way to develop systems. And scales well, since you can easily move into microservices.
@@MilanJovanovicTech How do you distinguish logical and physical? I would suggest there be a way to provide logical modules with less overhead and better scalability. At the moment it requires either under-documented, rather obscure Assembly files with multiple Assemblies defined in same file, or, more normally, separate projects compiled into separate DLLs.
