Operations | Monitoring | ITSM | DevOps | Cloud

Pop quiz: If you spend $40,000 per month on Anthropic, and you’ve got two customers, what’s your cost per customer? If you bypassed the easy answer of $20,000 and said, “Scott, you old trickster, that’s not enough information to answer that question,” you’ve won today’s prize: a lesson in the perils of average costs. Let’s flesh out the situation: You put an AI feature in your product, a document assistant powered by Claude.

Most growing teams eventually consolidate onto a full observability platform, and for teams correlating logs, metrics, and traces across a complex system, that’s often the right call. But a dedicated log tool still wins for a specific set of teams: ones that need to move fast, keep costs simple, and get real answers from logs without carrying the weight of a platform they don’t fully need yet. Here’s when that’s you.

Changing a Kubernetes resource request takes two numbers: what the workload requests, and what it uses. The CloudZero MCP server now returns both, by cluster, namespace, or workload. This gives you a number you can defend. Usage comes back as P95 over the date range you query, 30 days by default. When an engineering lead asks whether a service runs on a smaller request, that is the figure that settles it. Over-provisioning and under-provisioning show up on the same query.

The ability to log into all corporate applications with one username and password is pretty convenient, even compared to a password manager. As a benefit, the IT department can centrally enforce one desired two-factor auth mechanism. Now we, Icinga, also provide a so-called OpenID Connect integration for single sign-on. By the end of this text you’ll know how to connect your Icinga Web instance to the ID provider of your choice.

LLM token cost is the price a provider charges per token a model reads or writes, quoted in dollars per million tokens. Input and output bill at separate rates, with output priced at roughly 5x input. As of September 2026, published rates range from under $0.10 to more than $180 per million tokens on top-end reasoning tiers. In late 2025, Hardik Sonetta of Thomson Reuters Labs published a warning about the most common prompt caching mistake in production.

This blog post is part of PagerDuty’s ongoing series on how we’re helping customers navigate their journey towards autonomous operations. Read on to learn about how PD Automation Runner, now in Early Access, builds towards this vision. Your on-call engineer gets paged for a critical alert on a self-hosted Kubernetes cluster. If this were a cloud-hosted service, Workflow Actions and the SRE Agent would already be working the problem: pulling logs, checking health, surfacing a fix.

A team ships at a steady pace for most of the year. Then an audit approaches, and delivery slows. Engineers get pulled off feature work to support the audit, producing the configuration exports, logs, and environment checks that the evidence depends on. The slowdown lasts as long as the audit does. It is tempting to read this as a team that needs to move faster or be bigger. It is usually neither.

This blog post is part of PagerDuty’s ongoing series on how we’re helping customers navigate their journey towards autonomous operations. Read on to learn about how Scoped OAuth for Public Apps, now in Early Access, builds towards this vision. Your security team asks a simple question during a routine review: which third-party apps can reach our PagerDuty data right now, and what exactly can they do with it?

Table of Contents For most of Kubernetes’ history, memory management has been a blunt instrument. Cross your limit, and the kernel kills your container. There has been no equivalent to CPU throttling, no graceful backpressure, just a hard stop. With Kubernetes 1.37, that changes: Memory QoS, built on cgroups v2, graduates to Beta and is enabled by default.