9

We want to deploy multiple microservices on AWS ECS.

The issue that we need to solve is how to deploy them in an atomic way:

Let says we have the front-end services which requires the user service version 2.0.

How to make sure the user service is available before the front-end one is deployed.

We have the idea to replicate the whole infra, deploy everything then make it available. But it seems to me complex. How do you make these spare instances, deploy the services, and make the switch?

We are using two ALB and target groups for each services (dynamic port mappings).

Improve this question
1

3 Answers 3

Reset to default
8

Such question could be an indication of a poor architectural slicing into microservices. From What are Microservices?:

These services are built around business capabilities and independently deployable by fully automated deployment machinery.

The key point missed in such case would be their independently deployable aspect.

The point could also be simply misunderstood: it doesn't necessarily mean that each microservice should be capable to perform its entire functionality by itself, it just means that it's able to gracefully bail when some other microservice its functionality depends on isn't available, maybe keeping track of "todo" stuff for later execution, when that microservice becomes available.

Seen from a different perspective: handling unavailability of external service dependencies should actually be part of each microservice functionality - so one could consider the microservice fully functional even when its runtime dependencies aren't satisfied :)

It's worth noting, I think, that in the above comments I'm referring to microservices unavailability regardless of its reason, be it deployment dependency order or something else (outages, for example).

Another thing to consider would be the finer difference between deploying the microservices and bringing them into action by switching the traffic to them. Switching traffic is often faster than the deployment itself, so the availability at the entire system level can be increased by first deploying all microservices and only after that performing the traffic switch for all of them.

Improve this answer
3
  • That's what I thought! Thank you very much for these explanations.
    – Kaymaz
    Jul 23, 2018 at 12:16
  • 1
    The OP's situation may or may not be due to "poor architectural slicing". It is a realistic situation. The 2nd edition of "Building Microservices" concedes as much: "Domain coupling describes a situation in which one microservice needs to interact with another microservice, because the first microservice needs to make use of the functionality that the other microservice provides...In a microservice architecture, this type of interaction is largely unavoidable."
    – ottodidakt
    Dec 1, 2022 at 6:54
  • It's true that microservices should be deployable independently, but when microservice X requires microservice Y, and microservice Y is down, microservice X will typically provide limited set of functionalities (e.g. return no data or some old data from cache). Sometimes deploying multiple dependant microservices one after another makes sense, e.g. during creating a new environment (e.g. for new on-prem customer), or during creating temporary environment for end to end testing. And sadly I haven't seen a good solution to that.
    – iirekm
    May 11 at 12:39
1

We had a similar situation wherein one of our applications had to be built, deployed and be up & running before the second one was. This was because application A has some functionality that should be executed prior to it being referenced via application B. The way we accomplished this was by building it into a single delivery pipeline. That way, all the dependencies were deployed either sequentially or parallely as needed. We also had our database scripts in sequence along with rollback capabilities, just in case. Rewording to your case, you could deploy your api prior to your front-end service. Hope this helps.

Improve this answer
0
1

The solution here is to deploy the micro services using the blue/green deployment pattern.

Improve this answer

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.