Author
Kim
Backend software engineer working with Java, Spring Boot, and Kafka. I write about engineering problems, performance work, and architecture decisions from development and operations.
-
Kafka Consumer Lag: What the Number Measures, Why Your Client Metric Disagrees, and What Actually Moves It
Consumer lag measured three ways on Kafka 4.3.1 (CLI, Admin API, client metric), a 480 ms latency penalty from a popular fetch.min.bytes setting, and two ‘safety’ branches in a common poll loop that never execute.
-
Spring Boot GraphQL with Spring for GraphQL: Batch Loading, Error Shapes, and the Three Security Mistakes That Compile
A tested Spring Boot 3.5.16 bookstore API: measured N+1 with @SchemaMapping vs @BatchMapping, what validation errors actually look like to a client, and why @PreAuthorize on a controller can be doing nothing at all.
-
Java Virtual Threads on JDK 21: What 2,000 Connections, One Carrier Thread, and a synchronized Block Actually Do
A measured look at virtual threads on JDK 21.0.3: an echo server under 2,000 concurrent connections, the synchronized pinning stall reproduced with one carrier, where the 10x win is, and the case where virtual threads were slower.
-
Automating AI Model Deployment: A Pipeline You Can Run on kind, and Two Ways the Usual Workflow Silently Does Nothing
A train-build-push-rollout-verify-rollback pipeline run against a kind cluster and a local registry, with recorded output. It shows why a model image needs an immutable tag (three different models pushed as :latest gave a green pipeline and stale pods), why KUBECONFIG must be a path, and what in the Terraform and GitHub Actions parts could only…
-
Percentage Rollouts in Application Code: Bucketing, Cohort Correlation, and a Flag Metric With a Denominator
A tested Node.js flag evaluator with user, segment and percentage toggles: measured bucket distribution, why hashing the user id alone gives every flag the same cohort, and an evaluations counter whose exposition passes promtool.
-
Exactly-Once in Kafka Streams: One Crash, Two Guarantees, and Where the Transaction Stops
The same crash-and-restart run under exactly_once_v2 and at_least_once on Kafka 4.3.1: 100 correct outputs versus 159 with double counts, what a read_uncommitted consumer sees during the aborted transaction, and a side effect that ran 160 times for 100 inputs either way.
-
Streaming File Uploads in Spring Boot: 300 MB Under a 64 MB Heap, Measured in WebFlux and MVC
What a multipart upload actually does in Spring Boot 3.5.16: when the handler runs relative to the bytes on the wire, why a size check inside FilePart.content() cannot abort early, why the usual per-buffer APPEND write fails outright, and 300 MB uploads under -Xmx64m in both stacks.
-
Idempotency Keys in a Spring Boot REST API: Atomic Claims, Request Fingerprints, and 20 Concurrent Duplicates
A Spring Boot 3.5.16 payment endpoint with Idempotency-Key handling in Redis, tested with 20 simultaneous duplicates (one row, 1 x 201, 19 x 409), and the two ways the usual SETNX-lock recipe fails: a failed first attempt locks the key for a day, and the template it injects does not exist.
