Why 'Self-Healing' Infrastructure Requires Both Redundancy and Automated Failover

·1 views

The term 'self-healing' infrastructure is often used loosely to describe systems that simply have a restart policy. Genuine resilience depends on two elements working together: redundancy, which guarantees a healthy replica exists, and automated failover, which detects faults and reroutes traffic without human intervention. Redundancy without failover leaves spare capacity idle, while failover without redundancy has no healthy target to route to. The article also notes that these measures do not address data loss or faulty deployments, which require separate backup and disaster recovery planning with defined recovery time and recovery point objectives. It further argues that chaos engineering is needed to verify that failover actually works before a real outage exposes hidden flaws.

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)

Keeping postal codes as strings in Python fixture round trips

Postal codes can look numeric, but converting them to numbers drops leading zeros that cannot be restored by converting back to text. Codes with letters and spaces, such as Canadian-style formats, also need to be preserved exactly. A tutorial demonstrates an offline Python 3.10+ check using only the standard library to verify a JSON address export against a CSV file with a ZIP column. The check confirms that postal values remain strings, source references are retained, and each CSV row matches its JSON record field by field. The fixtures are invented for testing purposes and are not real postal reference data.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Developer fixes PWA sign-in bug in calorie tracker, asks readers to test the app

A developer building the ForkCount calorie tracker explains how they fixed a bug that broke the app when installed as a progressive web app, which involved changing login redirects and disabling URL session detection in Supabase. Despite 258 followers on dev.to, the developer reports zero app signups and asks readers why they haven't tried the free, card-free app. The post invites readers to sign up, describes the app's calorie target and BMR features, and asks users to report bugs with their device details. The developer notes that recent additions include photo logging, a barcode scanner, and streaks, while email verification and offline mode are still in progress. The post notes that the developer is 13 years old and built the app with Google AI Studio.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

AppConfig feature flags let Lambda teams roll out changes gradually with automatic rollback

AWS AppConfig feature flags allow switches in Lambda functions to change behavior without redeployment, applying updates in stages rather than to every function at once. The article contrasts this with Parameter Store, where a value change reaches all functions as soon as their cache expires, leaving no safety net for risky features. AppConfig adds schema validation, staged rollout strategies, and rollback triggered by CloudWatch alarms. The author argues Parameter Store suits harmless, rarely changing settings, while AppConfig suits switches that alter behavior. The walkthrough builds a cinema pricing API with CDK and TypeScript, including a gradual rollout and a kill switch.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Production Server Swapped in About 6 Minutes Using Scripted rsync Syncs

A developer moved a production e-commerce site from an Amazon Linux 2 server running PHP 8.2 to a new Amazon Linux 2023 instance on PHP 8.3 with php-fpm. The shared managed database meant only code, uploads, sessions, and admin-edited templates needed transferring. Data was synced in advance with rsync, using sudo on the sending side to read restricted session files, and a final sync ran during a maintenance window. Moving sessions let logged-in customers stay signed in through the cutover, and the maintenance page was up for roughly six minutes. The DNS change was kept as a manual step, with the old server left running as a rollback option.

0 comments Read more at DEV Community