Augustus O. (@_kekule15) header
Augustus O. (@_kekule15) profile picture

Software Engineer

Joined August 2019
Photo by Augustus O. (@_kekule15)Photo by Augustus O. (@_kekule15)Photo by Augustus O. (@_kekule15)Photo by Augustus O. (@_kekule15)Photo by Augustus O. (@_kekule15)Photo by Augustus O. (@_kekule15)Photo by Augustus O. (@_kekule15)Photo by Augustus O. (@_kekule15)Photo by Augustus O. (@_kekule15)Photo by Augustus O. (@_kekule15)
Backend Engineering Question: If XML and JSON can both represent structured data, why do most modern REST APIs use JSON? What advantages does JSON offer? And are there situations where XML would still be a better choice?
69
9
3
117
84,346
System Design Question: Design an API that accepts video generation requests which may take several minutes. The API must return immediately without waiting for the video. How would you handle: Job persistence? Worker execution? Status tracking? Retries and failures? Server restarts? Duplicate requests? Would you use a database-backed job system, a message broker or both?
2
1
5
System Design Question: Design a background job processing system that can survive server crashes and restarts. Jobs may take several minutes to complete and may depend on external providers. How would you: Persist job state? Recover unfinished jobs? Avoid duplicate submissions? Track external job IDs? Handle failed or abandoned jobs? Assume your application can restart at any time
Your API accepts a video-generation request and returns HTTP 202 with a job ID. Where do you store that job ID? If it's only in application memory, a server restart can leave you unable to track the operation. Persist the job state and provider reference so another application instance can recover and monitor it. Accepted jobs should not disappear when a server restarts
1
Senior Backend Interview Question: If database transactions can guarantee ACID properties, why can't we simply use them for everything in a microservices architecture? What exactly does the Saga pattern solve that traditional database transactions cannot?
2
1
12
Database concurrency: Most developers know about transactions. Fewer really understand the different isolation levels and what problems each one actually prevents. Dirty reads, non-repeatable reads, phantom reads… Which isolation level do you usually use in production, and have you ever had to change it because of a real bug?
2
One of the most dangerous assumptions in backend engineering is believing that a request will only be processed once. Networks fail. Clients retry. Queues redeliver messages. Webhooks get sent again. Design important operations to be idempotent. Your system should be able to survive repeated requests without repeating unintended side effects
5
I think developers sometimes introduce Redis, RabbitMQ, Kafka and Celery into projects before they even understand the actual problem they're trying to solve. Infrastructure is not an achievement. Knowing when NOT to introduce infrastructure is also software engineering
5
4
192
Everyone talks about Celery, Redis, RabbitMQ and distributed workers. But what if you're building an MVP that needs to process asynchronous jobs without introducing another infrastructure dependency? One possible approach : PostgreSQL stores job records. APScheduler triggers periodic polling. A worker queries pending jobs. Job status is updated in the database. Unfinished jobs are recovered after a restart. You don't always need a complicated infrastructure to solve a relatively simple problem
12
Your backend has 15 API endpoints. All 15 work perfectly. But your application can still fail because of one background worker that nobody thought to monitor. We spend so much time discussing API performance. Who is monitoring the jobs that run after the API has already returned a successful response?
6
2
5
338
Software engineers: What happens if a distributed lock expires while the original worker is still executing?
16
Distributed locks are supposed to guarantee that only one worker runs a critical section at a time. In practice, they don’t always. A long GC pause is enough to break the guarantee Here’s how it happens: ↓
19
Augustus O. retweeted
Your backend has 15 API endpoints. All 15 work perfectly. But your application can still fail because of one background worker that nobody thought to monitor. We spend so much time discussing API performance. Who is monitoring the jobs that run after the API has already returned a successful response?
6
2
5
338
A database transaction cannot magically make an external API call transactional. You can successfully update PostgreSQL, call a third-party API, and still end up with inconsistent states. Distributed systems are full of these uncomfortable realities
5
2
6
228
A database transaction can solve a consistency problem. A message queue can solve an availability problem. A cache can solve a latency problem. A load balancer can solve a distribution problem. Know the problem before choosing the tool
16
Software developers : Can a system have race conditions even when every database transaction is atomic?
14
Augustus O. retweeted
A database transaction cannot magically make an external API call transactional. You can successfully update PostgreSQL, call a third-party API, and still end up with inconsistent states. Distributed systems are full of these uncomfortable realities
5
2
6
228
Senior Backend Interview Question: What's the difference between a load balancer and an API gateway? Can an API gateway handle everything a load balancer does? If you're designing a production system from scratch, would you deploy both or start with just one?
4
2
1
7
608
A payment provider takes 90 seconds to confirm a transaction. Would you make your API wait for the entire process before responding? What happens if your gateway times out after 30 seconds but the provider eventually confirms the payment? The client timing out doesn't necessarily mean the operation failed. How would you design this interaction?
1
37
System Design Question: You already have an API Gateway in your architecture. Do you still need a Load Balancer?
1
43
Augustus O. retweeted
Senior Backend Interview Question: What's the difference between a load balancer and an API gateway? Can an API gateway handle everything a load balancer does? If you're designing a production system from scratch, would you deploy both or start with just one?
4
2
1
7
608