diff --git a/content/posts/removing-dependency-between-microservices/images/new-system.png b/content/posts/removing-dependency-between-microservices/images/new-system.png
new file mode 100644
index 0000000..d98fa4d
Binary files /dev/null and b/content/posts/removing-dependency-between-microservices/images/new-system.png differ
diff --git a/content/posts/removing-dependency-between-microservices/images/old-system.png b/content/posts/removing-dependency-between-microservices/images/old-system.png
new file mode 100644
index 0000000..62a533d
Binary files /dev/null and b/content/posts/removing-dependency-between-microservices/images/old-system.png differ
diff --git a/content/posts/removing-dependency-between-microservices/index.md b/content/posts/removing-dependency-between-microservices/index.md
new file mode 100644
index 0000000..7e02ac4
--- /dev/null
+++ b/content/posts/removing-dependency-between-microservices/index.md
@@ -0,0 +1,72 @@
++++
+date = "2022-08-23T10:00:00+07:00"
+author = "duchh"
+description = "A small approach we learned when migrating from monolithic to microservice"
+title = "Removing dependency between microservices"
+categories = ["Helm", "NATS-Jetstream"]
+tags = ["Helm", "NATS-Jetstream"]
+slug = "removing-dependency-between-microservices" 
++++
+
+***In this blog post, I will show for you guys solution to solve the problem we have met when deploying in microservices***
+
+#### Problem
+
+A little bit information about our system:
+
+**Helm**: It helps us install, upgrade, manage Kubernetes applications.
+
+**Postgres**: We are using postgres version 13.1 for our database.
+
+**Nats-jetstream**: We are using it for our message broker system. We also have a blog about 
+[it](https://blog.manabie.io/2022/03/set-up-nats-multi-tenant-in-golang/).
+
+Currently, our system is in process migrate from monolithic to microservice. We deploy our all services by one helm chart, their 
+deploy process are parallel. In order to start a service, it requires some logics like: `connect DB success, connect Nats success, 
+register logic, upsert streams of nats-jetstream, create subscriber of nats-jetstream,...` if one of these things fail then the deploy 
+process fail too. So what is `upsert streams of nats-jetstream, create subscriber of nats-jetstream` means :thinking:?
+
+1. **Upsert streams of nats-jetstream**: We write a library to check the `stream` by golang code. Is it exist? -> create a new one. 
+Is it have any update? -> update it. Is it need to delete? -> delete it.
+
+2. **Create subscriber of nats-jetstream**: We need a `subscriber` to listen messages are published from publisher then it will 
+process messages base on our logic.
+
+![alt](./images/old-system.png)
+
+
+Scenario: This is a simulation about our system. We have two services `Teacher and Student`, **Teacher** service has a `teacher-stream` 
+stream, **Student** service has a `student-sub` subscriber and this subscriber will subscribe `teacher-stream` to receive messages. 
+If `teacher-stream` is not exist then `student-sub` is never be created -> **Student** service is depend on **Teacher** service, 
+this architecture is really bad. So we decide need to refactor this architecture before it makes problem. As you can see the image 
+above, when we deploy our application but `Teacher` service is fail then `Student` is service fail too.
+
+#### Approach
+At the beginning, we have some solutions:
+1. We think we will create streams, consumers of nats-jetstream in one place, we can use [nack](https://github.com/nats-io/nack) 
+to do that, but we have some custom logic when we want to `create, update, delete` stream or consumer. So this tool isn't enough 
+useful with our use case.
+
+2. We try to add logic `upsert teacher-stream stream` into both **Teacher** and **Student** services. Whether `teacher-stream` 
+isn't created by **Teacher**, it can be created by **Student**. So we can guarantee **Student** is never crash 
+due to `teacher-stream` isn't exist. But reality, we have a lot of subscribers subscribe to `teacher-stream` from other services. 
+Another thing, we want only one service own that stream, because that stream should be only managed by one team, it has many configs, 
+if other can modify its config not owner, this thing can make another problem. 
+
+3. In this approach, we think about write a `job` to upsert stream, it solves two issues: we can use our library to work with 
+nats-jetstream when we upsert stream, we can do that logic in one place. But the question is how/when we run this `job` :thinking:? 
+The answer is we can use [helm hook](https://helm.sh/docs/topics/charts_hooks/). This approach is the best so we choose this 
+solution for our system.
+
+#### New architecture
+
+![alt](./images/new-system.png)
+
+As you can see the image above, we use `pre-hook` to run the `job` upsert stream first, if the hook runs fail then helm will fail 
+and stop before deploying new version. On the order hand, the hook runs success then helm continue deploy services. As you can see, 
+**Student** service is deployed success whether **Teacher** service is crashing. They are not dependent on together anymore 
+:dancers: :dancers: :dancers: .
+
+#### Conclusion
+We have many other ways to solve this problem as long as it suits your system. At Manabie, we still have a lot of things need to do to 
+migrate our system from monolithic to microservices. Thanks all of you read!!!
\ No newline at end of file