-
429 chaos experiments across 22 operators: what actually broke
I ran operator-chaos against 22 Kubernetes operators with 429 experiments. PodKill mostly passed. DeploymentScaleZero and NetworkPartition revealed the real gaps. Then I built chaos-instrumented operator images using the SDK and found 4 new bugs plus 2 RBAC gaps that CLI testing could never reach.
-
Chaos engineering for Kubernetes operators: testing what actually matters
Traditional chaos tools kill pods and watch them restart. But operators manage entire resource graphs, not just Deployments. I built operator-chaos to test the real question: does the operator detect and correct drift across everything it manages?
-
Why your RBAC linter misses privilege escalation chains
kube-linter catches wildcards and cluster-admin bindings. But it cannot detect indirect privilege escalation through binding chains. Here is what it misses and how kube-chainsaw fixes it.
-
Why your operator caches everything in the cluster (5 anti-patterns)
The Spark Operator OOMKill was not unique. Here are 5 anti-patterns that cause unfiltered informer caches in controller-runtime operators, including code paths invisible during review.
-
Protect your Kubernetes Operator from OOMKill
How an unfiltered informer cache lets any user with edit permissions crash your operator. I found this in the Spark Operator, but the pattern is everywhere.
-
Building an AI-powered browser extension for tab lifecycle management
How I built a Chrome extension with a skill-based AI integration that lets any AI agent manage browser tabs through natural language, using a bridge REST API and curl commands.