node_configsMySQL
node_idis_active
acme-corp1
beta-inc1
gamma-llc0
Reads your database
Point Lynq at a MySQL table. Every row where activate is true becomes a managed node — no pipelines, no glue code.
Skip to content Your Database.
Lifecycle
Infrastructure as Data
Your Database.
Your Infrastructure.
Lynq turns database records into Kubernetes resources. Automatically.
0resource types
0lifecycle policies
MySQLbuilt-in datasource
Kubernetes 1.28+cluster requirement
Apache 2.0open source license
Capabilities
Everything each row needs, handled
One database row in, a full set of reconciled Kubernetes resources out — with the controls to keep it safe.
Server-Side Apply
Resources are applied with SSA under the lynq field manager — Lynq owns exactly its fields and never clobbers the rest.
Dependency-aware ordering
Declare dependIds and Lynq builds a DAG, applying resources in topological order and waiting for readiness gates.
deployment:{{ .uid }}-app
service:{{ .uid }}-svc
ingress:{{ .uid }}-web
DeploymentServiceIngress
One template, every node
Write a LynqForm once. Your columns render into a full resource set per row — Deployments, Services, Ingresses, and more.
Every resource tracked
Each LynqNode reports ready, pending, failed, skipped, and conflicts — so you always know the real state, per row.
Watch a Row Become a Running App
One database row, reconciled end to end — Lynq reads it, creates a LynqNode, applies the resources, and the app goes live.
How It Works
From Data to Resources in Seconds
A real operator session — apply the manifests, insert a row, deactivate another, and watch Lynq create and garbage-collect the matching Kubernetes resources. Only your database changes.
The Problem
Your Database Knows. Your Cluster Doesn't.
The moment a row changes, the cluster is out of date — until someone runs kubectl. Lynq closes that gap continuously: the cluster is always a reflection of the database.
Safety & Control
Policies That Protect Your Cluster
Data-driven automation is only safe if it knows what not to touch. Every resource in a LynqForm carries three per-resource policies — each a guardrail against a specific way automation could do damage.
conflictPolicy
Another controller already owns this Service.
Stucksafe default
Halts and surfaces the conflict — overwrites nothing.
Forceopt-in
Takes ownership deliberately via SSA force=true.
deletionPolicy
A row is deactivated — is its data wiped with it?
Deletedefault
Removes the resource along with the row.
Retainkeeps data
Keeps the PVC — drops the ownerRef, adds an orphan marker.
creationPolicy
Every reconcile could re-run a one-time init Job.
WhenNeededdefault
Re-applies whenever the rendered spec drifts.
Onceruns once
Creates once, then never touches it again.
Rollout safety
A big update, without the thundering herd
A template edit, a bulk update, or a large insert can touch — or create — nodes across the whole graph at once, and doing them all together would stampede your API server. maxSkew caps how many change at a time. Pick a trigger, drag maxSkew, and watch it roll through the topology.
maxSkew4
YAML
web-stack.yaml editedevery node re-renders → all reconcile updating 0 / 4 updated 0 / 36 pending 36
A single template edit reconciles every node — a thundering herd that can melt the control plane if it all happens at once. maxSkew rolls it through a few at a time while the rest keep serving the current version. Drag maxSkew ↑ for a faster rollout, ↓ for a gentler one.
Observability
Catch Problems Before They Spread
15 Prometheus metrics, a pre-built Grafana dashboard, and 14 alert rules with runbooks. Whatever goes wrong — conflicts, failures, latency, a whole hub — the metric graphs it and the right alert pages you.
rate(lynqnode_conflicts_total[5m])Grafana · conflicts
alert if > 0.1/s for 10m
Alertmanager
LynqNodeNewConflictsDetectedinfo
New conflicts on acme-corp-web-stack — resource_kind=Service.
LynqNodeHighConflictRatewarning
0.23 conflicts/sec on Service · policy=Stuck. Review naming templates.
The metric caught the drift and paged before it spread — no resource was overwritten.
Dashboard
Observe the Full Hub → Form → Node Graph
A web UI that shows live resource health, reconciliation events, and topology relationships — no kubectl required
localhost:8080/topology
Topology ViewHub → Form → Node hierarchy with live status
Resource HealthReady, pending, failed counts per node
Event StreamReconciliation events and error details per node
Quick Search⌘K to find any hub, form, or node instantly
Blog
Latest from the Blog
Insights on Infrastructure as Data, Kubernetes operations, and Lynq updates
Is It Okay to Delegate Business Logic to Infrastructure?
Thoughts on delegating business logic to the infrastructure layer, through the lens of Uber embedding Rate Limiting into their service mesh
Read article →In the Age of AI Agents, How Should Infrastructure Change?
The era of AI Agents autonomously managing infrastructure is coming. But do we have an answer to the question, 'When AI decides, who executes?'
Read article →Introducing Lynq Dashboard
Lynq Dashboard is now available. A web UI to visualize Hub-Form-Node relationships and quickly identify problematic resources.
Read article →Start Automating Infrastructure from Your Database
Requires Kubernetes and cert-manager. The quickstart provisions a full local environment — MySQL, Lynq, and sample resources — using automated setup scripts.