Great video! Please consider creating a series of videos about MassTransit, with a focus on topics like State Machines, Routing Slips, and more. MassTransit is a fantastic library to be used within microservices!
I came up with the exact same logic of having a Service that returns a Result from a request before seeing this video. Love to see Milan coming to the same conclusions!
I was also very confused with MassTransit's weird way of dealing with polymorphism. This video cleared a lot of the confusion I had with the documentation.
It looks like it's synchronous because the caller has to "block" and wait for a response. But the benefit is that the services don't have to know about each other - unlike with HTTP. You need to know which API to call and where to find it.
great video, but I could not find this pattern described in MassTransit documentation. It's a pity you missed a part how you set up a message broker here. Did you use separate queue for request/response? Also I found that request/response is async pattern for http communication, while request/reply is async messaging pattern
Hi Milan, awesome content as always! Is it a good practice to completely remove the use of MediatR and switch to Request/Response with MassTransit (InMemory, perhaps)? I'm a bit confused, but don't both serve the same purpose? Could you please provide some insights on this? Thank you very much.
Consider the ease of use with both libraries. MediatR is much simpler to configure and use than MassTransit. If all I need is an in-memory bus, MediatR is still a great choice
You should probably handle timeouts for Requests as well because there is no guarantee you'll get a Response. This is one of the fun things about Event driven systems 🙂
The default timeout is 30s. If it takes more than that to get a response, you probably shouldn't be doing request-response anyway. But I agree that it should be handle in some way
You don't need to know the address of the other service - only the message contract. This can be a big advantage. The downside is increased latency, as we're passing around more messages.
the benefit of MT is that you are decoupled from the transport/bustype. so locally you can run rabbit, or even rabbit in a docker. and in the cloud you use azure servicebus or other cloud bus. in unittest you can run in memory bus. your code is transport agnostic, only in the startup of the bus you need to start one bustype depending on an appsetting.
The reason you said you register the ArticleViewsService as scoped is because IRequestClient is also scoped, how is it possible to generally check by DI container and the dependency chain to make sure I don't do anything funny in the registration I wasn't supposed to do? Is there a tool to see all the DI chain or the places where things are not using the same registration type?
if you use the built in DI you will find out for sure without tools, when a singleton tries to use a scoped service. the other way around is perfectly fine. You can use scoped services in a singleton by creating a scope around it.
@@jeroen7362 we had an issue with the DB context that was used by a singleton and we changed it to be transient to solve the problem, don't exactly remember. I wanted a tool that shows me the dependency tree to find these errors if I suspect I did something wrong with the service registration.
@@makemeafirewall as far as i have experienced it will always throw a runtime exception like "Cannot consume scoped service from singleton" This would usually only happen during development if you even touch the appservice in your debug session.
@@jeroen7362 i think it gets more complicated when a transient is injected with a scoped or the other way around. There should be a way to get the tree and search it, also it would be great to see the request pipeline because sometimes the project is already running in production and you are not always aware of the order of things, especially handlers that are used or even registered using reflection
Hello Milan, i use this request/response pattern with rabbitMQ, but sometimes i have an issue: when api B is in idle state (not recieving requests) rabbitMQ connection of api B is lost and when i send reqeust from api A with requestClinet i get timeoutexception. is there some method to overcome this issue?
@@MilanJovanovicTech ok but is there a way keep other side always available..to configure rabbitMQ connection to be all the time alive while api running..do u know such thing? If not, what is solution ..
I requested you that make a video on .NET Core, static file response compression. Pre-Compress (br / gzip) or run time by .NET Core or IIS. Which are best and how to implement
MassTransit + RabbitMQ adds many benefits. For example, imagine your dB is down. Rabbit will retry your request (within your defined parameters) for you
By the way I agree with the comment below, like using some BFF to orchestrate. And I think messaging Request-Response pattern is better suited to workflows long running business process that span multiple boundaries when we are dealing with sagas. Don't you agree?
Want to master Clean Architecture? Go here: bit.ly/3PupkOJ
Want to unlock Modular Monoliths? Go here: bit.ly/3SXlzSt
Great video! Please consider creating a series of videos about MassTransit, with a focus on topics like State Machines, Routing Slips, and more. MassTransit is a fantastic library to be used within microservices!
I already have a few about MassTransit, but I'll happily do more
I would like to see these as well
Hi have you ever compared with cap library? What the differences
I came up with the exact same logic of having a Service that returns a Result from a request before seeing this video. Love to see Milan coming to the same conclusions!
I was also very confused with MassTransit's weird way of dealing with polymorphism. This video cleared a lot of the confusion I had with the documentation.
Glad we're on the same page 😁
You always make the videos I most need. Thank you !
I have a hunch for that 😁
What is benefits of using RabbitMq instead of http is this case? It looks like synchronous way of communicating
Http is synchronous, rabbit mq request-response is async
It looks like it's synchronous because the caller has to "block" and wait for a response. But the benefit is that the services don't have to know about each other - unlike with HTTP. You need to know which API to call and where to find it.
Great video! Thanks so much!
Glad you enjoyed it!
Thanks a lot! Exactly what I just needed :)
You're welcome! :)
How does it correlate response properly , doesn’t seem to enable sessions on a queue?
Attaches a message ID
great video, but I could not find this pattern described in MassTransit documentation. It's a pity you missed a part how you set up a message broker here. Did you use separate queue for request/response? Also I found that request/response is async pattern for http communication, while request/reply is async messaging pattern
Here: masstransit.io/documentation/concepts/requests
Hello Milan,
Would you please consider making a course about Microservices?
I love so much your videos !
It will happen at some point :)
Hi Milan, awesome content as always! Is it a good practice to completely remove the use of MediatR and switch to Request/Response with MassTransit (InMemory, perhaps)? I'm a bit confused, but don't both serve the same purpose? Could you please provide some insights on this? Thank you very much.
Consider the ease of use with both libraries. MediatR is much simpler to configure and use than MassTransit. If all I need is an in-memory bus, MediatR is still a great choice
imnsho Mediatr is for intraprocess messaging. Masstransit / Rebus are for interprocess messaging.
You should probably handle timeouts for Requests as well because there is no guarantee you'll get a Response. This is one of the fun things about Event driven systems 🙂
The default timeout is 30s. If it takes more than that to get a response, you probably shouldn't be doing request-response anyway.
But I agree that it should be handle in some way
what advantage does this solution have over http request response?
You don't need to know the address of the other service - only the message contract. This can be a big advantage.
The downside is increased latency, as we're passing around more messages.
MassTransit+RabbitMq ，The consumer fails and can be put into the standby queue. What needs to be done?
Handle the error on the request client side
great video! How about making your next one on Circuit Breakers or Bulkhead pattern or more on MassTransit ? I'm sure it would be awesome
Great ideas!
Can we use mass transit if we are making this cloud native microservices ? Benefit of using mass transit ?
the benefit of MT is that you are decoupled from the transport/bustype. so locally you can run rabbit, or even rabbit in a docker. and in the cloud you use azure servicebus or other cloud bus. in unittest you can run in memory bus. your code is transport agnostic, only in the startup of the bus you need to start one bustype depending on an appsetting.
@@jeroen7362 thanks a lot
You absolutely can
@@MilanJovanovicTech thanks Milan so can you guide me the steps for that ? Or does it work internally as a container ?
Great video.
One doubt, when we use AsNoTracking in EF, we need to call it at beggining or end of query?
It won't matter.
How did you registered in the DI container IRequestClient?
You don't have to, it's automatically registered
The reason you said you register the ArticleViewsService as scoped is because IRequestClient is also scoped, how is it possible to generally check by DI container and the dependency chain to make sure I don't do anything funny in the registration I wasn't supposed to do?
Is there a tool to see all the DI chain or the places where things are not using the same registration type?
if you use the built in DI you will find out for sure without tools, when a singleton tries to use a scoped service. the other way around is perfectly fine. You can use scoped services in a singleton by creating a scope around it.
@@jeroen7362 what happens when a singleton tried to use a scoped service without creating a scope? Do I get an error? Always?
@@jeroen7362 we had an issue with the DB context that was used by a singleton and we changed it to be transient to solve the problem, don't exactly remember.
I wanted a tool that shows me the dependency tree to find these errors if I suspect I did something wrong with the service registration.
@@makemeafirewall as far as i have experienced it will always throw a runtime exception like "Cannot consume scoped service from singleton" This would usually only happen during development if you even touch the appservice in your debug session.
@@jeroen7362 i think it gets more complicated when a transient is injected with a scoped or the other way around.
There should be a way to get the tree and search it, also it would be great to see the request pipeline because sometimes the project is already running in production and you are not always aware of the order of things, especially handlers that are used or even registered using reflection
Excellent video @MilanJonovicTech, is it possible to get a tutorial about Kafka
I have an entire series on Kafka with MassTransit: ruclips.net/p/PLx8uyNNs1ri0RJ3hqwcDze6yAkrmK1QI5
Chris stealing the show 😁
Great Video. Where can I download the source code for this?
Hi, I share the source code on Patreon
Hello Milan, i use this request/response pattern with rabbitMQ, but sometimes i have an issue: when api B is in idle state (not recieving requests) rabbitMQ connection of api B is lost and when i send reqeust from api A with requestClinet i get timeoutexception. is there some method to overcome this issue?
Nope. That's called temporal coupling. In request-response we assume the other side is always available. But that might not always be the case.
@@MilanJovanovicTech ok but is there a way keep other side always available..to configure rabbitMQ connection to be all the time alive while api running..do u know such thing? If not, what is solution ..
I requested you that make a video on .NET Core, static file response compression. Pre-Compress (br / gzip) or run time by .NET Core or IIS. Which are best and how to implement
Ok
@@MilanJovanovicTech thanks 👍
Hi Milan, if possible please create a small project explains microservice architecture completely
A small project can't explain that
Where can I get complete source code?
Patreon: www.patreon.com/milanjovanovic
Github for this expample?
I share the source code on Patreon
Don't see any added benefit in this use case. One more failure point in the response cycle.
Anything can fail in distributed systems, I don't see that as a counter point
MassTransit + RabbitMQ adds many benefits. For example, imagine your dB is down. Rabbit will retry your request (within your defined parameters) for you
But I guess the UI composition technique is better in this scenario.
Perhaps
@@MilanJovanovicTech
Any plans to cover UI composition in your videos?
By the way I agree with the comment below, like using some BFF to orchestrate.
And I think messaging Request-Response pattern is better suited to workflows long running business process that span multiple boundaries when we are dealing with sagas. Don't you agree?
👋