I have a running setup of Percona MySQL 5.7. I would like to try out Percona XtraDB cluster. The current data size is ~500GB. From what I've observed (seen in a couple of YouTube videos: https://www.youtube.com/watch?v=uUVdte3wB3I ), I reckon that the cluster members do the initial sync from the currently serving master. However, in my case, syncing 500 GB of data would take ages, which I can't afford. On the other hand, I have a backup of the current data, taken using percona-innobackupex. Can I use this backup to seed the cluster members (like I would feed the slave while doing master-slave replication) so that the initial sync takes less time?
Add a comment
|
1 Answer
Yes you can.
- For do this first start XtraDB Cluster on current server.
- Then create an incremental backup from current full backup with --galera-info switch.(It's very important to use --galera-info)
- Install Percona server in second server
- Prepare the backup files (both backups.Full and Incremental) on second server
- Open xtrabackup_galera_info file in your full backup directory and copy uuid and seqno (for example e0d744f3-c88a-11e6-ad0d-6ed1e3c29e6d:0)
create grastate.dat file in full backup directory and add the following lines
version: 2.1
uuid: e0d744f3-c88a-11e6-ad0d-6ed1e3c29e6d
seqno: 0
cert_index:
Back up is ready to be copied over to the MySQL data directory.
Stop mysql service
Delete all contents in /var/lib/mysql
copy-back prepare backup
innobackupex --copy-back /restore/full
chown mysql directory to mysql.mysql
Config galera variables in my.cnf
Start mysql
You can find more information in this topic:
http://severalnines.com/blog/how-avoid-sst-when-adding-new-node-galera-cluster
