Their MySQL server were the finest, and their procedures are the same, including backup procedures. Yet, concepts that matched a local shop my not fit a global firm.
Going from Local to Global
Their current procedures include both binary backup and mysqldump of the server. Although both procedures are aimed to keep the service available while performing the backup, the bottom line is that the CPU is highly utilized and the service is poor.
There are two major backup methods (you can read more about them and others at Lenz Grimmer's presentation):
- mysqldump: This method extracts the database content into a flat file. This file includes all the DDL and DML SQL statements. Although this method is considered as a "hot backup", it tends to reduce the SLA and leave users unsatisfied. Moreover, if the database is not locked during the process, you may result an inconsistent backup.
- Binary files backup: This method requires stopping the MySQL server. On the other hand if you copy the binary file to a disk on the same physical server, it could be a very short downtime. In order to keep the files in a safe place, you should later on backup these copied files over the network. Please notice that this method has a another disadvantage: one of the binary files may be corrupted and you may not notice that till getting to the money time.
There are three possible locations:
- The Master Server: The safest since we know that data was not "disappeared" during replication. The significant drawback is service downtime...
- The Existing Slave Server: At the risk losing information that was not replicated correctly, we get a service with no downtime. Yet, notice that if you use this server for other tasks you may affect these tasks SLA.
- Dedicated Slave: This can be a lightweight machine (yes, even a VPS) that can perform backups once or more times a day without causing downtime and without affecting the service at all.
Well, in this case, I suggested the customer choosing the Dedicated Slave option that in a very small price helped him reduce the service downtime and provide great SLA to their global customers.