Skip to content

Other Issues

Error Encountered when Creating Database with CR

The database is running normally, but an error occurs when creating a database using CR. This issue can be caused by the presence of special characters in the mysql root password.

image

  1. Retrieve and view the original password:

    [root@master-01 ~]$ kubectl get secret -n mcamel-system mcamel-common-mysql-cluster-secret -o=jsonpath='{.data.ROOT_PASSWORD}' | base64 -d
    
  2. If the password contains special characters such as - , attempting to enter the original password in MySQL Shell will result in the following error:

    bash-4.4# mysql -uroot -p
    Enter password:
    ERROR 1045 (28000): Access denied for user 'root'@'localhost' (using password: YES)
    
  3. To resolve this issue, you are offered two methods to clean up and rebuild:

    • Method 1: Clear the data directory and delete the Pod. Wait for the sidecar to be running, then delete the data directory again and repete deleting the Pod:

      [root@master-01 ~]# kubectl exec -it mcamel-common-mysql-cluster-mysql-1 -n mcamel-system -c sidecar -- /bin/sh
      sh-4.4# cd /var/lib/mysql
      sh-4.4# ls | xargs rm -rf
      
    • Method 2: Delete the PVC first, then delete the Pod to restore functionality:

      kubectl delete pvc data-mcamel-common-mysql-cluster-mysql-1 -n mcamel-system
      
      kubectl delete pod mcamel-common-mysql-cluster-mysql-1 -n mcamel-system
      

Note

Using the above methods to clean and rebuild will result in the database being reset and data loss.

The Following MySQL Message in the DCE 5.0 Platform

When the following message appears during operations in the management platform, it indicates that the master-replica relationship of the MySQL nodes has changed. However, other modules of the platform have not timely updated the connection objects, resulting in a write operation being executed on a read-only replica node.

The MySQL server is running with the read-only option so it cannot execute this statement

Solution: Go to the Container Management platform and restart all related replica nodes.

Comments