Galera Cluster - Diffrent DB size on each node / Resync Node

Comments

1 comment

  • Avatar
    Ashraf Sharif

    Hi,

    Regarding GRAheader for 5.7, you may try this one: https://www.dropbox.com/s/xeokdy4ok2cyc3o/GRA_header57.tar.gz?dl=0

    Regarding mysqlcheck:

    AFAIK, mysqlcheck command is not replicated in Galera and it uses the unsupported explicit LOCK/UNLOCK TABLE statements during the operation. Details on this can be found at https://mariadb.com/kb/en/library/mariadb-galera-cluster-known-limitations/ . The error log that you've shown was related to an error when MySQL was starting up, not the actual reason why the node got kicked out from the cluster at the first place. If you could give the whole MySQL error log, that would be helpful.

    Regarding node resync:

    Yes, that is the best way to fully resync a node in Galera, by wiping out /var/lib/mysql content (or simply delete grastate.dat). Depending on the DB size, it could cause performance impact to the donor node so handle it with care.

    Regarding adding users to new nodes:

    When a node rejoins and resyncs via SST, the new node should have all the users from the synced node. If you are adding a user via CREATE/GRANT statements, the user will be synced by Galera automatically. Not sure what you really mean by manually add users to new node, one use case I can think of if you add a user via 'INSERT INTO' statement to mysql.user table, which requires FLUSH PRIVILEGES and won't be synced by Galera due to its MyISAM storage engine.

    Regards,
    Ashraf

Please sign in to leave a comment.

Powered by Zendesk