Skip to content

Latest commit

 

History

History
82 lines (70 loc) · 2.97 KB

Lab_3_backup.md

File metadata and controls

82 lines (70 loc) · 2.97 KB
Raw

Back to Agenda

Lab 3 - Backup a MySQL Cluster

Backup and restore of NDB Cluster can be done centralized by MySQL Cluster Manager. NDB also have native backup and restore for community edition of NDB, more details here.

As NDB Cluster is a distributed database, the backup is also a distributed where each datanode will backup it's part of the complete dataset. Important that you make sure all parts of the backup are stored in a safe manner.

Local data/metadata in MySQL API nodes are not part of NDB backup, this data must be handled separately. This data can be secured by using mysqldump separately.

Attention:

Remember to set the PATH env varaible manualy or by using bash>. ./setenv before trying to start mcm client, mcm client need to find the mysql client.

Backup and Recovery

Start the mcm client If you want to see help run ./mcm/bin/mcm --help Our reference manual can be found here.

./mcm/bin/mcm

Run a full backup of the cluster:

mcm> backup cluster --waitstarted mycluster;

If you have 2 mcm windows open you can track progress by running:

mcm> show status --progress mycluster;

Once the backup is done use list backups to list stored backups:

mcm> list backups mycluster;

Log into MySQL client:

mysql -uroot -proot -h127.0.0.1 -P3310

And create some data:

mysql> create database lab;
mysql> use lab;
mysql> create table t1 (i int) engine=ndbcluster;
mysql> insert into t1 values (1),(2),(3);

Lets create a backup (in mcm client) to make sure we can recover in case of disaster

mcm> backup cluster mycluster;

Disaster strikes!

mysql -uroot -proot -h127.0.0.1 -P3310 -se"drop database lab"

Lets recover from last backup, first look at available backups and pick the last one, note the BackupId

mcm> list backups mycluster;

Before we can restore data we need to make sure we get current cluster into a clean state:

mcm> stop cluster mycluster;
mcm> start cluster --initial --skip-init=50,51 mycluster;
     (--skip-init is to skip initialization for mysql nodes, make sure NodeID's are correct)

Next step is to use recover the data using BackupId from "list backups" statement

mcm> restore cluster --backupid=2 --background mycluster;

During recovery we can follow the progress by looking at:

mcm> show status --progressbar mycluster;

If you want to re-run recovery just run the Disaster statement over again ;)

If we do some analysis before recovery and verify that only one bad SQL statement was executed (that dropped a table or a database), we can do a partial restore like:

mcm> restore cluster --include-databases=lab --backupid=2 mycluster;

Above partitial restore can be done without having to stop/start the cluster.

Back to Agenda