Retry Storm: How A Single User Crashed 30 ECS Tasks At Production
Infrastructure-level retry amplification, 503 ambiguity and the controls that protect a service from a failure multiplier.
Read on MediumTechnical notes
Ideas worth writing down before the next incident makes them urgent.
Practical, vendor-neutral notes on distributed systems, backend reliability, Go performance, Kafka, Redis and search infrastructure.
Published writing
Production stories and technical notes.
External articles link to Medium. Site notes are clearly marked as educational drafts and do not disclose confidential implementation details.
When Apache Solr’s Replication Handler Went Rogue
A production debugging story about replication, deletion policy timing and leadership changes.
Read on MediumRetry amplification: when “resilience” becomes the outage
A vendor-neutral companion note on retry ownership, deadlines, budgets and degraded behaviour.
Read site noteCache boundaries are an API design decision
Cache ownership, invalidation and degraded reads should be explicit parts of a service contract.
Read site noteTopics on the editorial roadmap
Systems are easier to improve when the vocabulary is shared.
Maintainable publishing
How new articles are added.
Article source material lives in content/articles/, with frontmatter for title, description, category, dates and reading time. The static HTML route is deliberately simple to keep the site fast and readable without a client-side content system. See the content editing guide in the repository before publishing a new post.