TLDR DevOps 2025-04-30

97% of performance issues aren't detected until they affect your users. But what if you could see failures developing at the time of release, and stop them before they impact users?

→ LaunchDarkly's Guarded Releases embeds real-time observability directly into your release pipeline. It monitors, detects, and mitigates failures automatically before they reach production.

Watch the webinar to see:

• Why today's release methods are failing
• How real-time visibility at the moment of deployment changes everything
• A live demo of Guarded Releases catching and preventing a failure in real time.
• A free Quickstart Guide to help you issue releases with confidence.

Access the recording →

OpenTofu has officially joined the Cloud Native Computing Foundation after receiving a license exception to continue using MPL 2.0.

A major power outage hit Portugal, Spain, and parts of France on April 28, disrupting transportation, businesses, and services, with Portugal's grid operator blaming "induced atmospheric vibration" from extreme temperatures in Spain. Cloudflare observed the outage's impact on Internet traffic, network quality, and routing across local, national, and network levels.

Kubernetes v1.33 now allows configurable tolerance for horizontal pod autoscaling, an alpha feature that lets users fine-tune the sensitivity of replica adjustments. The new feature allows different tolerances for scale-up and scale-down events to be set. This replaces the previous cluster-wide default tolerance of 10%, which was often too coarse for large deployments.

Azure will retire default outbound access for virtual machines on September 30, requiring new VMs to use explicit outbound connectivity methods like Azure NAT Gateway or a Public IP address. Existing VMs will continue functioning but should transition to an explicit method for better control and security, with multiple options available for implementing the change.

Talos Linux, an OS designed for Kubernetes, can be installed on pre-configured servers using the kexec utility to load the Talos Linux image from an existing Linux distribution.

A recent State of Platform Engineering Report found that 44.67% of organizations surveyed do not measure any metrics to prove the success of their platform engineering initiatives. Metrics for platform engineering success can include deployment frequency, lead time, failure rate, recovery time, Net Promoter Score (NPS), and the time it takes for new employees to create their tenth pull request.

Incident management focuses on quickly resolving disruptions, while problem management identifies root causes to prevent future issues. Combining both processes enhances system reliability, minimizes downtime, and fosters a proactive, continuous improvement approach.

If you want your free software to stay accessible with minimal upkeep, the best option in 2025 is to ship it as static HTML and JavaScript via GitHub Pages.

Meta's Llama 4 Scout 17B and Llama 4 Maverick 17B are now available on Amazon Bedrock as a fully managed, serverless option, offering multimodal capabilities with early fusion technology for image grounding and extended context processing,

Pulumi ESC now integrates with Infisical, an open-source secrets management platform, to centralize secrets and configuration management.