Intermediate concepts in DevOps & Cloud, page 12
Operator SDK: Build Kubernetes Operators Faster
The Operator SDK is a developer toolkit that scaffolds the boilerplate for building, testing, and packaging Kubernetes Operators. Use it to automate complex application lifecycle management, like deploying a database cluster that can self-heal and perform…
App of Apps Pattern: Manage Application Fleets, Not Individuals
The App of Apps pattern uses a single parent Argo CD application to declaratively manage a fleet of child applications. This is ideal for bootstrapping entire environments from one Git repo.
Incident Timeline Reconstruction
Timeline reconstruction is the after-the-fact process of merging evidence from logs, metrics, deploys, and chat into one accurate sequence of what happened.
Kubernetes Finalizers: The 'Do Not Delete Yet' Lock
A finalizer is a pre-deletion lock. It tells Kubernetes to block an object's deletion until a controller cleans up external resources, like a cloud database or storage bucket.
Argo Rollouts: Beyond Kubernetes Rolling Updates
Argo Rollouts replaces Kubernetes' basic RollingUpdate with safer, controlled strategies like canary and blue-green. It gradually shifts traffic to new versions while monitoring metrics, automating rollbacks if KPIs degrade.
Postmortem Report Template
A postmortem template is a standard structure for documenting an incident: summary, impact, timeline, root cause, what went well and poorly, and action items.
Flux HelmRelease: Declarative Helm Charts in Git
A Flux HelmRelease is like putting your helm install command into a YAML file and checking it into Git. It lets you declaratively manage Helm charts for any application via GitOps.
Actionable Remediation Items
Actionable remediation items are specific, owned, trackable follow-ups from a postmortem that prevent recurrence or improve response. Good ones are concrete and assigned with a due date; vague items like be more careful are non-actionable and predictably…
OLM: Kubernetes' App Store for Operators
OLM is the app store for Kubernetes Operators: it installs, updates, and resolves dependencies declaratively. Use it when managing third-party or custom Operators across clusters.
MLOps: Applying DevOps to Machine Learning
MLOps applies DevOps principles to automate the machine learning lifecycle, creating an assembly line for models. It's for moving from notebooks to production systems that retrain automatically.
Counterfactual Reasoning in Incident Analysis
Counterfactual reasoning asks what would have prevented or mitigated an incident: if this alert had existed, if this check had run. Used well it finds systemic gaps, but it is a trap when it implies a single person should have just acted differently in…
Feature Store: The Single Source of Truth for ML
A feature store is the single source of truth for ML models, acting as a central kitchen for prepped ingredients (features). It's used to ensure the same feature logic is applied in both training and real-time inference, preventing model drift.
The Second Story of an Incident
The first story blames human error and stops there; the second story asks why the action made sense to the person at the time and what systemic conditions enabled it. Seeking the second story is the heart of blameless, learning-oriented incident analysis.
Hyperparameter Tuning for LLM Inference
Control an LLM's creativity versus predictability by tweaking its inference parameters. This is crucial for tasks like generating structured JSON versus creative text. The footgun is changing parameters without a clear goal, leading to chaotic output.
Review of Reviews (Postmortem Metrics)
A meta-process that audits the quality and follow-through of postmortems themselves, tracking metrics like action-item completion, time-to-close, and recurrence of incidents.
ML Inference Endpoint: The API for Your Model
An ML inference endpoint is the stable API URL your application calls to get predictions. It separates the public URL from the underlying model, letting you swap models without changing client code.
Performance Profiling
Profiling measures where a program actually spends its time and resources, attributing CPU cycles, memory, or wall-clock latency to specific functions or call paths. It replaces guesswork with data so optimization effort targets the real bottleneck.
Backpressure
Backpressure is a mechanism by which a slow consumer signals an upstream producer to slow down or stop, preventing unbounded queues and resource exhaustion. It keeps systems stable under overload by propagating capacity limits backward through a pipeline.
LitmusChaos
LitmusChaos is an open-source, Kubernetes-native chaos engineering platform that runs fault experiments as custom resources. It injects failures like pod kills, network latency, and resource stress to validate that services stay resilient under real-world…
Hiring for SRE
Hiring for SRE seeks engineers who blend software-development skill with systems and operations depth, plus strong debugging and incident temperament. The hardest part is finding people who can both write automation and reason about failure at scale under…
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