Modernizing URL Rewriting Architecture with High Instance Count and Stability Issues

Question

Hi Team,

I'm working with a company that's experiencing challenges with their current architecture. Here's the scenario:

Current Setup:
- 150 individual t2.micro EC2 instances running Nginx as reverse proxies
- Each instance serves one customer for URL rewriting/shortening
- Instances distributed across us-east-1 (100) and sa-east-1 (50)
- Backend application runs on Oracle Cloud (APEX platform)

Main Issues:
1. Random instance unavailability (15-minute periods)
2. High costs with public IPs
3. No effective monitoring/alerting
4. Manual instance restarts required
5. Poor observability (Zabbix monitoring fails during outages)

Questions:
1. What's the recommended approach to modernize this architecture while maintaining URL rewriting capabilities per customer?
2. Are there any specific patterns or anti-patterns we should consider?

Thanks in advance!

Brettski-AWS · Accepted Answer

If it were me (and bear in mind that there are a lot of unanswered questions here) I would be doing the following:

1. Use a CloudFront distribution per customer.
2. If possible, do the redirects/rewrites in CloudFront using Lambda@Edge or CloudFront functions.
3. Don't run t2 instance types. Instead, create an ALB in each region (maybe more than one depending on load) and have a fleet of instances that can all deal with every customer. Use a more appropriate (higher performance) instance type. But less will be required and they can run "hot" rather than having a bunch of smaller instances that are probably running idle or maybe even being overloaded.
4. (3) will resolve the EIP billing problem.
5. Make sure (3) uses auto-scaling.
6. Use CloudWatch Metrics and Logs and Log Insights to try and determine where problems are occuring.

Gary Mclean · Answer

Agree with Brettski, however you still have other options as always.

As its NGINX, you can put all the same rules on 1 server NGINX.

* Create an ALB, with at least 2 EC2s OR ECS Containers running NGINX in a target group.
* Point all URLS to the ALB
* The NGINX Hosts still will perform URL rewrites etc.

Modernizing URL Rewriting Architecture with High Instance Count and Stability Issues

Relevant content