Customer Issue with AWS Redis

Question

I have a customer which recently go live with their Mobile platform on AWS and uses Redis extensively, this app has its FE on AWS and BE on premise. The FE APIs uses Microsoft Technology and IIS to run. The Application has been very unstable and we currently over-provision Servers to get some level of stability. Imagine running A mobile applications on 80 Servers on AWS for FE alone. Some components of the apps are still failing as I write. What is the best practice when accessing Redis from an External Microsoft Tech intensive app? How should the code be written to use the minimal servers on AWS and get some stability.

Didier_Durand EXPERT
9 months ago
Hi, you have to add tangible details to your question: metrics, error logs, etc. if you want to obtain meaningful support from re:Post community. "Very unstable" can mean millions of things: detailing in more details what is exactly failing will definitely help. Thanks

Answer

One critical code component that many people overlook is retries in exception handling.
There is a legacy approach that either something will succeed or it will fail; and that if it fails, it will continue to fail due to some system being hard-down, ie. the database. There are tons of reasons for transient errors, such as a DB lock, or a time-out due to resources that are in the process of auto-scaling.

It is critical to assume a non-zero error rate for legacy as well as modern, complex systems.

When transitioning from on-premises to the cloud, the underlying infrastructure gets abstracted and therefore even more complex. This complexity provides tremendous value including vastly more scalability and resiliency but the trade-offs include even more likelihood of non-zero error rates.
Having simple yet thorough exception handling as well as observability is complex but essential.

Customer Issue with AWS Redis

Relevant content