架构

部署过程

1、CRD
2、Operator Deployment
3、InnodbCluster CR

实例

apiVersion: v1
kind: Secret
metadata:
  name: mysql
stringData:
  rootUser: root
  rootHost: '%'
  rootPassword: 123456
---
apiVersion: mysql.oracle.com/v2
kind: InnoDBCluster
metadata:
  name: mysql
spec:
  secretName: mysql
  imageRepository: hub.geovis.com.cn/library
  tlsUseSelfSigned: true
  instances: 3
  router:
    instances: 3
  datadirVolumeClaimTemplate:
    accessModes:
      - ReadWriteOnce
    resources:
      requests:
        storage: 100Gi
  mycnf: |
    [mysqld]
    authentication_policy=mysql_native_password
    skip-host-cache
    skip-name-resolve
    datadir=/var/lib/mysql
    secure-file-priv=/var/lib/mysql-files
    user=mysql
    pid-file=/var/run/mysqld/mysqld.pid
    explicit_defaults_for_timestamp
    max_allowed_packet=16M
    bind-address=0.0.0.0
    character-set-server=UTF8MB4
    collation-server=utf8mb4_unicode_ci
    lower_case_table_names=1
    max_connections=100000
    wait_timeout=86400
    innodb_thread_sleep_delay=10000
    interactive_timeout=7200
    default-time_zone='+8:00'

查看 operator 日志

kubectl -n mysql-operator logs -f $(kubectl get pod -n mysql-operator -l name=mysql-operator -o jsonpath='{.items[0].metadata.name}')

查看 innodbcluster 状态

kubectl get innodbcluster --watch

删除出问题的 Pod (尽量避免直接删除 Pod)

kubectl delete pod mysql-cluster-2
kubectl delete pod mysql-cluster-2 --force --grace-period=0
kubectl patch pod/mysql-cluster-2 -p '{"metadata":{"finalizers":[]}}' --type=merge

尝试重启 router

kubectl -n mysql-operator delete pod -l name=mysql-operator

Service

不同的 Service 有不同的用途

{cluster-name}.{namespace}.svc.{cluster-name} 为 ClusterIP，负载到所有 Router，用于集群访问
{cluster-name}-instances.{namespace}.svc.{cluster-name} 为 Headless Service，负载到所有 MySQL-Server 的 Pod，用于管理、监控等用途

端口

InnodbCluster 使用不同端口实现读写分离等

mysql:            3306
mysqlx:           33060
mysql-alternate:  6446
mysqlx-alternate: 6448
mysql-ro:         6446
mysql-rx:         6447
mysqlx-ro:        6448
mysqlx-rw:        6449