Kestra 2.0 Launches With Rewritten Engine and Decoupled Worker Architecture

·2 views

Workflow orchestration platform Kestra has released version 2.0, featuring a major rewrite of its core engine four years after its initial public introduction. The update splits the system into a control plane and a data plane, allowing workers to run anywhere — including on-premises, in separate cloud regions, or inside outbound-only networks — without requiring database credentials or inbound connections. Workers now communicate with the control plane via a single persistent gRPC stream, with support for TLS encryption and certificate or JWT-based authentication. The release also consolidates what were previously two separate engine implementations into one unified executor and scheduler, reducing duplicated maintenance overhead. Kestra 2.0 remains licensed under Apache 2.0, continuing the project's open-source commitment.

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)

XRP Ledger Batch Transactions Pass QA: 164 Tests Cover Atomic Multi-Step Flows

A QA test report dated September 8, 2026, documents results for Batch (XLS-56) Transactions on the XRP Ledger, prepared by sgramkumar in a GitLab CI environment running on xrpld devnet. The feature allows accounts to bundle up to eight inner transactions into a single atomic unit, eliminating the risk of partial completion that existed when transactions settled independently. Testing covered the BatchV1_1 amendment, which tightened BatchSigner authorization and signing semantics over the original Batch amendment. A total of 164 tests were executed across three categories: 106 core functional tests, 56 adversarial and security tests, and 2 cross-feature interaction tests involving Permission Delegation, Sponsored Fees, and Confidential MPT. All four execution modes — AllOrNothing, OnlyOne, UntilFailure, and Independent — along with multi-account signing, fee rules, and metadata reporting were validated against the XLS-56 specification.

0 comments Read more at DEV Community

ProgrammingHacker News ·

Interactive Topological Picture Book Brings Abstract Math to Life Online

A digital project titled 'A Topological Picture Book, Rendered' has been published at e-infinity.space. The work appears to be an online visual rendering inspired by geometric and topological concepts. It was shared on Hacker News, where it received early attention from the community. The project offers an accessible, visual approach to exploring complex mathematical ideas. Limited discussion has taken place so far, suggesting it is a newly surfaced resource.

0 comments Read more at Hacker News

ProgrammingDEV Community ·

How to Deploy Your AI-Built App: A Practical Guide for First-Time Builders

Developers who build apps using AI assistants often hit a wall when trying to share their work beyond a local environment, because deployment involves infrastructure decisions that code generation tools do not handle. The article distinguishes between static front-end apps, which can be hosted for free on platforms like Vercel, Netlify, or Cloudflare Pages within minutes, and backend apps, which are better served by managed platforms like Railway, Render, or Fly before considering a dedicated server. A self-managed server is recommended only when platform costs exceed a basic VPS, or when persistent storage, background jobs, or data residency requirements arise. AI-generated code commonly embeds API keys directly in source files and relies on local file storage, both of which cause serious problems in production environments — including security breaches and data loss on redeployment. Developers are advised to use environment variables for secrets, managed databases instead of SQLite, and object storage services like Cloudflare R2 or Amazon S3 for user-uploaded files.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

JavaScript Promises: A Cleaner Fix for Callback Hell in Async Code

Asynchronous JavaScript is commonly handled using callbacks, but deeply nested callbacks — known as callback hell or the Pyramid of Doom — make code hard to read, debug, and maintain. JavaScript Promises offer a cleaner alternative by representing the eventual result of an asynchronous operation through three states: pending, fulfilled, or rejected. Once a Promise settles into either fulfilled or rejected, its state cannot change again. Developers use .then() to handle successful results, .catch() to manage errors, and .finally() to run code regardless of outcome. Promise chaining allows multiple asynchronous steps to be written sequentially, significantly improving code readability and error handling compared to nested callbacks.

0 comments Read more at DEV Community