1. 为什么要在Kubernetes上运行MySQL当第一次听说要在Kubernetes上部署MySQL时我和大多数DBA一样充满疑虑数据库这种有状态服务真的适合跑在容器环境里吗经过三年多的生产实践我可以负责任地说——不仅可行而且在特定场景下优势明显。Kubernetes为MySQL带来的核心价值在于弹性扩展能力传统MySQL主从架构扩容需要手动配置复制关系而Kubernetes的StatefulSet配合Operator可以实现自动化扩缩容资源利用率提升通过request/limit机制多个MySQL实例可以更精细地共享物理资源标准化部署流程将MySQL的配置、备份、监控等操作全部代码化消除环境差异故障自愈能力结合liveness/readiness探针可以实现节点故障自动转移但要注意不是所有场景都适合超大规模OLTP系统如每秒数万写入仍建议使用物理机或云数据库对网络延迟极其敏感的应用需要考虑Kubernetes网络插件性能没有持久化存储经验的团队需要谨慎评估存储方案2. 生产级MySQL on Kubernetes架构设计2.1 基础拓扑方案典型的Kubernetes MySQL部署采用主从复制架构每个MySQL实例包含一个StatefulSet Pod建议使用Percona Server for MySQL一个专属PVC通常用RWO存储类一个匹配的Headless Service一个用于读流量的ClusterIP Service# 示例StatefulSet片段 apiVersion: apps/v1 kind: StatefulSet metadata: name: mysql spec: serviceName: mysql replicas: 3 template: spec: containers: - name: mysql image: percona:8.0 volumeMounts: - name: data mountPath: /var/lib/mysql volumeClaimTemplates: - metadata: name: data spec: accessModes: [ ReadWriteOnce ] resources: requests: storage: 100Gi2.2 关键组件选型建议存储方案对比存储类型适用场景性能表现成本注意事项本地SSD PV高性能OLTP★★★★★高需处理节点亲和性云厂商块存储通用场景★★★★☆中注意IOPS限制Ceph RBD混合云环境★★★☆☆低网络延迟可能影响性能NFS开发测试环境★★☆☆☆极低不推荐用于生产环境网络插件选择Calico适合大多数场景提供网络策略控制Cilium高性能支持eBPF加速云厂商CNI与底层网络深度集成3. 部署实操从零搭建MySQL集群3.1 环境准备清单在开始前确保准备好Kubernetes集群v1.20默认StorageClass配置正确Helm 3已安装至少3个worker节点每个节点8C16G重要提示生产环境务必禁用swap否则可能引发OOM Killer误杀MySQL进程3.2 使用Bitnami MySQL Helm Chart部署对于大多数用户推荐使用Bitnami维护的MySQL Charthelm repo add bitnami https://charts.bitnami.com/bitnami helm install mysql-cluster bitnami/mysql \ --set auth.rootPasswordYourStrongPassword \ --set architecturereplication \ --set secondary.replicaCount2 \ --set persistence.size100Gi部署完成后验证状态kubectl get pods -l app.kubernetes.io/instancemysql-cluster # 预期看到1个master和2个replica3.3 自定义配置技巧通过ConfigMap挂载my.cnfapiVersion: v1 kind: ConfigMap metadata: name: mysql-config data: my.cnf: | [mysqld] innodb_buffer_pool_size4G innodb_log_file_size1G max_connections500然后在StatefulSet中挂载volumeMounts: - name: config mountPath: /etc/mysql/conf.d/my.cnf subPath: my.cnf volumes: - name: config configMap: name: mysql-config4. 运维关键监控、备份与故障处理4.1 监控方案实施推荐使用Prometheus Operator监控MySQL部署exporter sidecarcontainers: - name: mysqld-exporter image: prom/mysqld-exporter:v0.14.0 env: - name: DATA_SOURCE_NAME value: exporter:password(localhost:3306)/创建ServiceMonitorapiVersion: monitoring.coreos.com/v1 kind: ServiceMonitor metadata: labels: app: mysql-monitor name: mysql-monitor spec: endpoints: - port: exporter-port selector: matchLabels: app: mysql关键监控指标mysql_global_status_uptime运行时间mysql_global_status_threads_connected连接数mysql_global_status_innodb_row_lock_time_avg平均行锁等待4.2 备份策略设计采用定期快照binlog备份的组合方案使用Velero进行PVC快照velero backup create mysql-backup \ --include-resourcespods,persistentvolumeclaims \ --selector app.kubernetes.io/instancemysql-cluster配置cronjob执行逻辑备份apiVersion: batch/v1beta1 kind: CronJob metadata: name: mysql-dump spec: schedule: 0 2 * * * jobTemplate: spec: template: spec: containers: - name: dump image: mysql:8.0 command: [sh, -c] args: - mysqldump -h mysql-primary -u root -p$PASSWORD --all-databases | gzip /backup/dump_$(date %Y%m%d).sql.gz volumeMounts: - name: backup-volume mountPath: /backup4.3 常见故障处理案例1脑裂问题现象两个节点同时认为自己是master 解决方案检查quorum配置验证网络分区情况使用pt-heartbeat检测复制延迟案例2存储空间不足现象Pod不断重启日志显示No space left 应急处理kubectl exec -it mysql-0 -- mysql -e PURGE BINARY LOGS BEFORE NOW() - INTERVAL 3 DAY; kubectl scale statefulset mysql --replicas0 kubectl edit pvc mysql-data-mysql-0 # 调整存储大小 kubectl scale statefulset mysql --replicas35. 性能优化实战经验5.1 参数调优指南根据实例规格调整关键参数内存大小innodb_buffer_pool_sizeinnodb_log_file_sizetable_open_cache8GB4G512M200016GB12G1G400032GB24G2G8000通过压力测试验证配置sysbench oltp_read_write \ --db-drivermysql \ --mysql-hostmysql-primary \ --mysql-userroot \ --mysql-passwordyourpassword \ --tables10 \ --table-size1000000 \ --threads32 \ --time300 \ prepare5.2 读写分离实现使用ProxySQL实现智能路由部署ProxySQL集群apiVersion: apps/v1 kind: Deployment metadata: name: proxysql spec: replicas: 2 template: spec: containers: - name: proxysql image: proxysql/proxysql:2.4 ports: - containerPort: 6033配置路由规则INSERT INTO mysql_servers(hostgroup_id,hostname,port) VALUES (10,mysql-primary,3306), (20,mysql-replica-0,3306), (20,mysql-replica-1,3306); INSERT INTO mysql_query_rules (rule_id,active,match_pattern,destination_hostgroup,apply) VALUES (1,1,^SELECT.*FOR UPDATE,10,1), (2,1,^SELECT,20,1), (3,1,^INSERT,10,1);6. 安全加固最佳实践6.1 网络隔离方案使用NetworkPolicy限制访问apiVersion: networking.k8s.io/v1 kind: NetworkPolicy metadata: name: mysql-allow spec: podSelector: matchLabels: app: mysql policyTypes: - Ingress ingress: - from: - podSelector: matchLabels: app: application ports: - protocol: TCP port: 33066.2 证书管理使用cert-manager自动签发TLS证书apiVersion: cert-manager.io/v1 kind: Certificate metadata: name: mysql-cert spec: secretName: mysql-tls duration: 2160h # 90d renewBefore: 360h # 15d issuerRef: name: letsencrypt-prod kind: ClusterIssuer dnsNames: - mysql.internal.example.com然后在MySQL配置中启用SSLargs: - --ssl-ca/etc/mysql/certs/ca.crt - --ssl-cert/etc/mysql/certs/tls.crt - --ssl-key/etc/mysql/certs/tls.key7. 迁移传统MySQL到Kubernetes7.1 在线迁移方案使用pt-online-schema-change实现最小停机时间迁移在目标集群创建空实例配置主从关系CHANGE MASTER TO MASTER_HOSTlegacy-mysql, MASTER_USERrepl, MASTER_PASSWORDpassword, MASTER_AUTO_POSITION1;启动复制START SLAVE;验证数据一致性后切换应用连接7.2 数据一致性验证使用pt-table-checksum检查差异pt-table-checksum \ --replicatetest.checksums \ --databasesproduction \ hlegacy-mysql,uroot,ppassword pt-table-sync \ --replicatetest.checksums \ hk8s-mysql,uroot,ppassword \ hlegacy-mysql,uroot,ppassword \ --print8. 真实生产案例分享某电商平台迁移经验集群规模6个分片每个分片1主2从挑战黑五期间峰值QPS达到15k解决方案使用本地NVMe存储调整内核参数vm.swappiness1配置HPA基于CPU和连接数自动扩容成果平均延迟降低40%运维人力减少60%踩过的坑曾因默认StorageClass使用NFS导致性能瓶颈未设置Pod反亲和性导致两个主实例调度到同一节点忘记配置PDB导致维护时意外中断