Advanced everything in DevOps & Cloud
Datadog Infrastructure Basic targets host health checks
Datadog describes a lighter monitoring tier for host metrics and selected processes on noncritical workloads.
CNCF argues agents should propose system state, not receive production root
A CNCF Ambassador argues that agents should express changes as versioned system state and pass them through review, build and deployment controls instead of receiving unrestricted production root.
CNCF announces its 2026–2028 Ambassador cohort
CNCF accepted 58 new ambassadors for its 2026–2028 cohort, while 48 previous ambassadors stepped away.
ArgoCon previews the direction of Argo CD 4.0
CNCF previews ArgoCon North America 2026 and says the Argo community is beginning the visioning process for Argo CD 4.0. The event covers four distinct projects—Argo CD, Workflows, Rollouts, and Events—so learners can map each project to a delivery problem before treating the vision as a release.
AWS introduces CloudWatch Omni for collaborative observability
AWS announced CloudWatch Omni on 22 September, an OpenTelemetry-based experience for application and AI-agent observability outside the AWS console.
Kubernetes 1.37 adds a native unused-PVC condition
Kubernetes 1.37 promotes PVC unused tracking to Beta, adding an Unused condition managed by the PVC protection controller.
Docker publishes Sandbox Kit Specification v3
Docker published an open-source Kit Specification v3 for packaging an agent’s runtime, tools, permissions, and network policy with its environment.
GitHub describes an agentic fuzzing taskflow
GitHub Security Lab describes a taskflow that finds entry points, writes harnesses, runs AFL++, and triages crashes for C/C++ projects.
AWS adds shared, ordered EventBridge custom buses
AWS announced enhanced EventBridge custom event buses that can be shared across accounts, support ordered delivery, and use a Subscriber resource.
Grafana 13.2 adds multiple notification policy trees
Grafana says multiple notification policies are generally available in Grafana-managed alerting with Grafana 13.2, enabling independently managed routing trees.
GitLab publishes critical patch releases 19.4.1, 19.3.3 and 19.2.7
GitLab released 19.4.1, 19.3.3 and 19.2.7 on 23 September with important bug and security fixes for self-managed installations.
Docker introduces Cloud Sandboxes for long-running agents
Docker announced Cloud Sandboxes on 24 September: the same microVM isolation model as local Sandboxes, running on Docker-managed compute.
Critical vm2 sandbox escape (CVSS 10.0) traced to its own README example
GitLab's Threat Research Group found a CVSS 3.1: 10.0 sandbox escape in vm2, a widely used Node.js sandbox library, in a config copied from vm2's own README. Version 3.11.7 fixes the attack, but the underlying risk remains for anyone using require.external. Update immediately and prefer containers for untrusted code.
Designing an error budget policy
Define SLO and budget, tiered consequences as burn worsens, a feature freeze on exhaustion, and concrete earn-back criteria.
Client-side chaos for an uncontrollable third party?
Inject faults at your client boundary via a proxy or fault-injecting wrapper, simulate timeouts, errors, and latency, then verify timeouts, retries, breakers, and fallbacks.
Chaos test for gray-failure cascades in shared services?
Inject partial latency into a shared service, hypothesize tenants stay isolated within steady state, and monitor cross-system queue depth, pool saturation, retries, and per-tenant SLIs.
Reliability paved roads for an internal PaaS?
Built-in observability, safe deploys with health checks and rollback, sane timeouts/retries/limits, and SLO tooling.
Design a chaos experiment for a payment dependency?
Define a measurable steady state, hypothesize it holds when payments fail, limit blast radius to a small traffic slice, and auto-abort on SLO breach.
Replication and consistency for active-active regions?
Choose per data class between synchronous (low RPO, higher latency) and async replication, address write conflicts, and reason via CAP and PACELC.
When horizontal scaling is the wrong fix
Adding instances fails when the bottleneck is a shared resource like one database, a lock, or a queue, so more instances just add contention; investigate where time is actually spent.
We are hiring for this. Every open role lists the topics its interview covers, so you can prepare for the real thing rather than guessing.
See open roles