As we know, big applications can be difficult to manage. To tackle this problem, we use microservices, in which we break down a large application into smaller and independent parts. Each microservice performs a particular task and has its own database. Now, all of these microservices will interact with each other to make the system work smoothly. We will understand how to tackle these inconsistent state issues in this post.
For example, we all have ordered pizza online. Have you ever noticed that there are many different processes involved in ordering a pizza?
Ordering Service – When ordering a pizza, we first use the ordering service.
Payment Service – The second thing we use is the payment service to pay for our pizza.
Delivery Service – The last thing is the delivery service where the delivery service assigns a delivery person. So that it can deliver your pizza to you.
when the Pizza ordering process is divided into many small processes. Making these processes into subparts will make the process easy to handle and it will work smoothly.
To learn about the microservices its official website.
What is an Inconsistent State?
As we know, Everything comes with its ups and downs. Similarly using microservices also has its ups and downs. As we have already discussed in a microservices system, each service performs a specific task and possesses its database. However, these microservices are not connected completely due to this, they can have different or outdated data. This difference in data can lead to an Inconsistent State.
For example,
You have ordered a pizza, and the order service confirms it. The payment service processes the payment.
But due to some miscommunication, the restaurant didn’t receive your order, due to this the restaurant will never know that you have ordered a pizza and they need to prepare it, and since they have not prepared it you will not be able to get your pizza.
This can happen due to an inconsistent state in which some services are unable to communicate or update the shared data to complete the process and make the system work smoothly…
Why do these Inconsistent state issues happen? Does it happen on their own or out of the blue?
Why Do Inconsistent State Issues Happen?
The microservices system works when different services communicate with each other over a network, in which every one of these services has its databases. But sometimes due to some issues, the services fail to communicate with each other which leads to an inconsistent state. Now let’s understand what are these issues and why they are caused in the first place.
What are these issues?
- Network delays: Microservices are like humans they talk to each other over the internet, just like websites that connect to their backends. What is there is a small delay in communication, Hence the services might get old or the data might be incorrect which can cause problems.
- Service Failures: You might have come into the occurrence when something is downloading or an update stops out of random
because of some stupid error. The same thing happens in microservices. If a single service crashes while working, some tasks may not finish which can lead to missing data. - Separate Databases (Database Isolation): Every service has its own database, and they do not directly share their data with each other. If services get disconnected or are not able to sync their data, the system may show that it has outdated or wrong information.
In a microservices architecture, traditional database transactions are impossible because each service has its own database. this is where we use SAGA patterns. SAGA pattern allows us to make smaller parts across multiple services while maintaining consistency
How to Fix Inconsistent State Issues?
If any step in a multi-service transaction fails, SAGA ensures that the system doesn’t end up in an inconsistent state. So all the previous steps are either rolled back or compensated to maintain a consistent state
Different types of SAGA
Choreography- Choreography is a self-managed process in which each service acts on its own with no central service control. It is good for workflows with few steps
Orchestration- It is a central controlling process in which a central service manages the entire transaction which tells each service what to do. It is good for workflows with many dependencies
Ensuring Actions Are Idempotent
What is idempotency?
If we repeat the same action multiple times and the result remains the same this is said to be idempotency. without this, the customer could be charged twice
How to Prevent Duplicate Actions:
Use unique request IDs- Every request gets its ID so that the system checks if the ID has already been processed.
Keep Track of Completed Requests- If the same request comes in again, the system checks whether the request is been processed or not, if the request has already been processed the system ignores it.
Use Reliable Messaging to Keep Services in Sync– In most cases, microservices work with each other but sometimes these messages can get lost or delayed, causing the data to go out of sync.
Keeping Microservices in Sync
Microservices send messages to each other, but sometimes these messages get lost or delayed which can cause data to mismatch.
How to fix this?
- Message Queues: If a service is temporarily down, a queue stores messages and delivers them later.
- Event Sourcing: The system records a history of all events.
Keeping Data Fresh and Correct
Using outdated data can lead you to the wrong or incorrect decisions which can further lead you to errors.
How do we keep data up to date?
- Get data from the right place: When you fetch or get data instead of using old data.
- Use Fast storage: Storing frequently used data in the cache to reduce load and speed up the system.
- Keep Backup Copies of Data: Saving backup copies of important data to restore the information if something goes wrong
Monitoring and Detecting Issues Early
If we needed the system to run smoothly, we had to find and fix problems before they could cause major failures.
Some Tools to Detech issues:
- Logging Systems: We need to store a particular record of errors so that we can identify the problems quickly
- Alerting Tools: These tools send us notifications instantly when something goes wrong so engineers can fix it before it can affect users.
Conclusion
Like humans, Microservices talk to each other by sending messages to each other, but sometimes these messages take time to arrive. This can cause major issues. For Example, If you pay for an order, but the order system does not get the update immediately your order might still show as “Pending” even though you have already paid.
Related Bugs and Fixes Post:- AI Model Serverless Deployment? Read This Before You Regret!
