Dev Rebuilds CI/CD Pipeline to Eliminate Silent Production Failures on VPS

·1 views

A developer overhauled their Node.js/TypeScript deployment architecture after discovering a critical blind spot: remote scripts ran without strict error handling, causing GitHub Actions to report pipeline success even when the production application had crashed. The original setup triggered deployments on every merge to the main branch via SSH, git pull, and pm2 restart, with database migrations running before the application build — meaning a TypeScript error could leave the database altered with no new code deployed. Since Prisma lacks native down migrations, reverting such failures required high-risk manual intervention. The redesigned system requires an annotated SemVer tag to trigger any deployment, passing through six sequential gates including tag validation, automated testing with Vitest, a live migration rehearsal on a disposable copy of the production database, and post-deploy smoke tests. The developer has published the full modular GitHub Actions template as an open-source repository for teams managing similar VPS-hosted applications.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Study of 4,000 Experiments Finds Simple Retraining Schedules Beat Smarter ML Methods

A study by researcher Sawan Dasari ran nearly 4,000 experiments to evaluate how and when machine learning teams should retrain deployed models experiencing concept drift. The research compared four retraining strategies — no retraining, fixed periodic schedules, error-threshold triggers, and statistical drift detection — and found that for models lacking incremental learning, the choice of policy can swing accuracy by 15 to 55 percentage points. Counterintuitively, simple periodic retraining outperformed more sophisticated reactive approaches under both abrupt and gradual data drift. The study also identified a latency-budget interaction where compute constraints and retraining time compound together, potentially cutting a team's effective retraining capacity by half without them realizing it. The findings suggest most production ML teams are over-engineering their retraining triggers while underestimating the value of predictable, scheduled updates.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Task vs. Ownership: Why How You Delegate Work Changes Everything

There is a meaningful distinction between assigning someone a task and giving them true ownership of a problem. A task requires completing a defined action, while ownership demands understanding the goal, making decisions, and taking responsibility for the outcome. Ownership does not mean abandoning someone, especially if they are new — context, checkpoints, and guidance remain important. Effective ownership must also come with sufficient authority, since holding someone accountable without letting them make decisions is not genuine ownership. The broader goal of good delegation is to progressively distribute problems and decision-making, not just workload.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Anthropic exits computer-use beta, OpenAI adds zero data retention for API users

Anthropic moved its computer-use tool out of beta on August 19, 2026, replacing it with a new toolset that requires developers to update their request structure rather than simply swapping a header. A new browser automation toolset was also released alongside it, giving the Claude API a first-party lane for driving in-browser applications. On the same date, OpenAI announced Zero Data Retention for eligible API customers, meaning prompts and responses will not be stored after processing, along with a preview of Private Safety Processing that monitors interactions without exposing raw content to OpenAI. The safety feature is currently in limited testing and is slated for broader rollout in September 2026. Separately, MCP core maintainers published an updated roadmap on August 22 outlining five workstreams, including agent identity standards and server-initiated messaging, signalling where the protocol is headed after its stateless-core rewrite in July.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Practical SEO Tips: Core Web Vitals, Metadata, and Accessibility Explained

A developer and freelancer has shared a practical guide to improving search engine optimisation, drawing on years of hands-on experience with personal and client projects. The guide emphasises that Google's primary goal is to answer user queries, making genuine, well-structured content with natural keyword placement more effective than outdated tactics. It highlights Core Web Vitals — LCP, INP, and CLS — as real-user ranking signals since 2021, with 2026 thresholds remaining unchanged, and recommends tools like PageSpeed Insights and Chrome's Lighthouse tab for measurement. Proper metadata setup, including unique title tags, meta descriptions between 150–160 characters, canonical tags, and Open Graph or Twitter Card tags, is outlined as essential for both search and social media visibility. The guide also draws a strong connection between web accessibility practices — such as semantic HTML, descriptive alt text, and adequate touch targets — and better search engine comprehension of a page.

0 comments Read more at DEV Community