How Kraken Migrated 8M Daily RabbitMQ Messages to v4 With Zero Downtime

·1 views

Engineering teams at Kraken faced a high-stakes upgrade from RabbitMQ v3.13 to v4.2.2, which introduced breaking changes including the removal of Global QoS and Classic Queue Mirroring. With 8 million messages processed daily and strict uptime requirements, the team could not afford service interruptions or data loss during the transition. Their strategy involved spinning up a parallel virtual host running quorum queues bound to topic exchanges, while using a feature flag to gradually shift application pods from the old setup to the new one. Celery's Native Delayed Delivery feature, which requires quorum queues, was leveraged to preserve ETA-based task execution that would otherwise have blocked workers under v4. The migration was executed in rolling phases, ensuring both environments ran simultaneously until all messages and workers were fully transferred to the new configuration.

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)

WireGuard GitHub Action Verifies Tunnel Health and Auto-Cleans CI Connections

A new open-source GitHub Action, ankurk91/wireguard-action, lets CI runners connect to private networks via WireGuard VPN for the duration of a job, eliminating the need to whitelist GitHub IP ranges or maintain self-hosted runners. Unlike existing solutions, it actively verifies the tunnel handshake after connection, failing the step immediately if the peer is unreachable rather than letting jobs silently time out later. The action also automatically tears down the interface and deletes the config file after each run, regardless of whether the job succeeds, fails, or is cancelled. Built with no external dependencies and a minimal Node.js shim over auditable bash, it requires only an Ubuntu runner and an IPv4-only WireGuard config stored as a repository secret. Optional diagnostics mode and a troubleshooting guide address common CI-specific pitfalls, including the absence of IPv6 on GitHub runners and routing conflicts caused by full-tunnel AllowedIPs settings.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

x402 Protocol Enables AI Agents to Make Autonomous Micropayments via HTTP

The x402 specification repurposes HTTP's long-dormant 402 status code to create a payment negotiation channel between AI agents and paid API services. When a server requires payment, it returns a 402 response with details such as amount, token, blockchain, and a nonce to prevent replay attacks. The client then signs and submits an on-chain ERC-20 transaction, attaching the transaction hash in a retry request header. Because the entire flow operates within standard HTTP headers and response bodies, no new protocol layer is required. Developers can implement x402 using existing tools like the viem library for Ethereum wallet interactions and standard HTTP clients.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

n8n Releases AI Security Monitoring Guide for LLM-Powered Workflows

Automation platform n8n has published a guide on securing AI workflows in production environments, warning that traditional monitoring tools are insufficient for LLM-based systems. The guide highlights risks such as prompt injection, adversarial inputs, data poisoning, supply chain vulnerabilities, and model drift, each of which leaves distinct signals that standard infrastructure monitoring can miss. Unlike conventional software, AI workflows can remain technically operational while producing unsafe, incorrect, or out-of-baseline outputs. n8n recommends a layered observability approach combining model-level telemetry — including inputs, outputs, confidence scores, and access patterns — with anomaly detection to establish behavioral baselines. The guidance is aimed at security and engineering teams using AI to automate tasks like document summarization, customer request routing, and agent-based tool use.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Client-Side PDF to Markdown Conversion Offers Privacy and Performance Gains

Converting PDFs to Markdown improves compatibility with modern developer workflows, including Large Language Model (LLM) pipelines and static site generators like Docusaurus and Hugo. Unlike server-side converters that upload files to remote infrastructure, browser-native conversion keeps documents entirely on the user's machine, ensuring zero data transmission and built-in privacy compliance. Client-side processing also eliminates backend infrastructure costs and delivers faster results by leveraging the user's local CPU and GPU via browser Web Workers. Developers can implement this using Mozilla's PDF.js for rendering and the Turndown library to convert extracted HTML content into structured Markdown. The approach also enables Git-based version control of converted documents, making change tracking straightforward through standard pull requests.

0 comments Read more at DEV Community