feat(deploy): 增加三节点 K3s 高可用迁移能力

新增 WireGuard 全互联、三 server embedded-etcd K3s、CloudNativePG 双实例、Barman OSS 备份、双 NGINX、Kubernetes Secret/RBAC 与本地旧文件按严格 24 小时清理。

新增维护窗口数据迁移、digest 固定滚动发布、应用回滚、跨节点文件 E2E、节点和数据库故障演练、CNPG 恢复与 etcd 快照验收脚本;洛杉矶仅作为带 NoSchedule 污点的仲裁节点。

所有生产 Secret 只在执行时从 0600 本地环境和旧生产容器导入,仓库不保存凭据;公网入口保持人工 DNS 故障切换边界。

验证:bash -n、ShellCheck、kubectl kustomize、Node 语法检查、Secret 扫描、OSS put/head/delete 实测。
This commit is contained in:
2026-07-28 04:37:31 +08:00
parent edf7e66941
commit 971540a2a4
38 changed files with 3386 additions and 3 deletions
+113
View File
@@ -0,0 +1,113 @@
apiVersion: barmancloud.cnpg.io/v1
kind: ObjectStore
metadata:
name: easyai-postgres-oss
labels:
app.kubernetes.io/part-of: easyai-ai-gateway
spec:
retentionPolicy: 14d
configuration:
destinationPath: s3://wangbo0808/easyai-ai-gateway/production/postgresql
endpointURL: https://oss-cn-shanghai.aliyuncs.com
s3Credentials:
accessKeyId:
name: easyai-oss-backup
key: ACCESS_KEY_ID
secretAccessKey:
name: easyai-oss-backup
key: ACCESS_SECRET_KEY
wal:
compression: gzip
maxParallel: 2
data:
compression: gzip
jobs: 2
instanceSidecarConfiguration:
env:
- name: AWS_DEFAULT_REGION
value: oss-cn-shanghai
- name: AWS_REQUEST_CHECKSUM_CALCULATION
value: when_required
- name: AWS_RESPONSE_CHECKSUM_VALIDATION
value: when_required
---
apiVersion: postgresql.cnpg.io/v1
kind: Cluster
metadata:
name: easyai-postgres
labels:
app.kubernetes.io/name: easyai-postgres
app.kubernetes.io/part-of: easyai-ai-gateway
spec:
instances: 2
imageName: ghcr.io/cloudnative-pg/postgresql:18.4-standard-trixie@sha256:4587df73024408f5b2be9b4dd6ba2ccee8c9e5dc0c9a87c274c292291cc8a68c
imagePullPolicy: IfNotPresent
primaryUpdateStrategy: unsupervised
primaryUpdateMethod: switchover
failoverDelay: 0
switchoverDelay: 60
stopDelay: 90
smartShutdownTimeout: 30
bootstrap:
initdb:
database: easyai_ai_gateway
owner: easyai
secret:
name: easyai-postgres-app
postInitSQL:
- CREATE EXTENSION IF NOT EXISTS pgcrypto;
storage:
storageClass: local-path
size: 20Gi
walStorage:
storageClass: local-path
size: 5Gi
affinity:
enablePodAntiAffinity: true
podAntiAffinityType: required
topologyKey: easyai.io/site
nodeSelector:
easyai.io/database: "true"
resources:
requests:
cpu: 500m
memory: 1Gi
limits:
cpu: "2"
memory: 3Gi
postgresql:
synchronous:
method: any
number: 1
dataDurability: preferred
parameters:
archive_timeout: 60s
max_connections: "200"
shared_buffers: 512MB
wal_compression: "on"
wal_keep_size: 1GB
monitoring:
enablePodMonitor: false
plugins:
- name: barman-cloud.cloudnative-pg.io
isWALArchiver: true
parameters:
barmanObjectName: easyai-postgres-oss
---
apiVersion: postgresql.cnpg.io/v1
kind: ScheduledBackup
metadata:
name: easyai-postgres-daily
labels:
app.kubernetes.io/part-of: easyai-ai-gateway
spec:
# Six-field cron, UTC: 18:30 UTC is 02:30 Asia/Shanghai.
schedule: "0 30 18 * * *"
backupOwnerReference: self
cluster:
name: easyai-postgres
method: plugin
pluginConfiguration:
name: barman-cloud.cloudnative-pg.io
immediate: false
suspend: false