cleanup

k8s/README-migrations.md

@@ -1,69 +0,0 @@
-# Database Migration Strategy for Kubernetes
-
-This document describes how database migrations are handled in the Kubernetes deployment to ensure only one node runs migrations at a time.
-
-## Overview
-
-In a distributed environment with multiple pods, we need to ensure that database migrations only run once, not multiple times concurrently. We use two strategies:
-
-1. **PostgreSQL Advisory Locks** - For runtime migration coordination
-2. **Init Containers** - For deployment-time migrations (recommended)
-
-## Implementation
-
-### 1. PostgreSQL Advisory Locks
-
-The `Aprsme.MigrationLock` module uses PostgreSQL advisory locks to ensure only one node can run migrations at a time:
-
-- Uses `pg_try_advisory_lock()` to acquire a non-blocking lock
-- Other nodes wait for the lock to be released
-- Automatically releases the lock after migrations complete
-
-### 2. Init Container (Recommended)
-
-The StatefulSet can be configured with an init container that runs migrations before the main pods start:
-
-```bash
-# Apply the init container patch
-kubectl patch statefulset aprs -n aprs --patch-file k8s/statefulset-init-container-patch.yaml
-```
-
-This approach:
-- Runs migrations sequentially before any pods start
-- Prevents race conditions
-- Makes migration failures visible in pod events
-
-### 3. Auto-Migration Disabled in Cluster Mode
-
-When `CLUSTER_ENABLED=true`, automatic migrations on startup are disabled to prevent race conditions.
-
-## Manual Migration
-
-To run migrations manually:
-
-```bash
-# Run on the first pod
-kubectl exec -it aprs-0 -n aprs -- /app/bin/migrate
-
-# Or create a one-off job
-kubectl run migrate-job --rm -it --image=ghcr.io/aprsme/aprs.me:latest \
-  --env="DATABASE_URL=$DATABASE_URL" \
-  --env="SECRET_KEY_BASE=$SECRET_KEY_BASE" \
-  --env="MIX_ENV=prod" \
-  --env="SKIP_DB_CREATE=true" \
-  --restart=Never \
-  -- /app/bin/migrate
-```
-
-## Configuration
-
-Environment variables:
-- `CLUSTER_ENABLED=true` - Enables distributed locking
-- `SKIP_DB_CREATE=true` - Skips database creation (for PgBouncer)
-
-## Best Practices
-
-1. **Use Init Containers** for production deployments
-2. **Test migrations** in a staging environment first
-3. **Monitor migration logs** during deployments
-4. **Have rollback plan** ready with `mix ecto.rollback`

k8s/poddisruptionbudget.yaml

@@ -1,11 +0,0 @@
-apiVersion: policy/v1
-kind: PodDisruptionBudget
-metadata:
-  name: aprs-pdb
-  namespace: aprs
-spec:
-  minAvailable: 1
-  selector:
-    matchLabels:
-      app: aprs
-  unhealthyPodEvictionPolicy: IfHealthyBudget

k8s/statefulset-deployed-at-patch.yaml

@@ -1,13 +0,0 @@
-apiVersion: apps/v1
-kind: StatefulSet
-metadata:
-  name: aprs
-  namespace: aprs
-spec:
-  template:
-    spec:
-      containers:
-      - name: aprs
-        env:
-        - name: DEPLOYED_AT
-          value: "TIMESTAMP_PLACEHOLDER"

k8s/statefulset-init-container-patch.yaml

@@ -1,37 +0,0 @@
-apiVersion: apps/v1
-kind: StatefulSet
-metadata:
-  name: aprs
-  namespace: aprs
-spec:
-  template:
-    spec:
-      initContainers:
-      - name: migrate
-        image: ghcr.io/aprsme/aprs.me:latest
-        imagePullPolicy: Always
-        command: ["/app/bin/migrate"]
-        env:
-        - name: DATABASE_URL
-          valueFrom:
-            secretKeyRef:
-              key: database-url-pgbouncer
-              name: aprs-secret
-        - name: SECRET_KEY_BASE
-          valueFrom:
-            secretKeyRef:
-              key: secret-key-base
-              name: aprs-secret
-        - name: MIX_ENV
-          value: prod
-        - name: SKIP_DB_CREATE
-          value: "true"
-        - name: CLUSTER_ENABLED
-          value: "false"  # Disable clustering for migration runner
-        resources:
-          limits:
-            cpu: 200m
-            memory: 256Mi
-          requests:
-            cpu: 100m
-            memory: 128Mi

k8s/statefulset-patch.yaml

@@ -1,12 +0,0 @@
-apiVersion: apps/v1
-kind: StatefulSet
-metadata:
-  name: aprs
-  namespace: aprs
-spec:
-  updateStrategy:
-    type: RollingUpdate
-    rollingUpdate:
-      partition: 0
-      maxUnavailable: 0  # This ensures pods are only terminated after new ones are ready
-  minReadySeconds: 30  # Wait 30 seconds after a pod is ready before considering the update successful

k8s/statefulset-startup-probe-patch.yaml

@@ -1,19 +0,0 @@
-apiVersion: apps/v1
-kind: StatefulSet
-metadata:
-  name: aprs
-  namespace: aprs
-spec:
-  template:
-    spec:
-      containers:
-      - name: aprs
-        startupProbe:
-          httpGet:
-            path: /health
-            port: 4000
-          initialDelaySeconds: 10
-          periodSeconds: 5
-          timeoutSeconds: 3
-          successThreshold: 1
-          failureThreshold: 12  # Gives 60 seconds total to start up