This talk will cover forecasting and planning future growth for high volume mysql instances occupying many (500+) servers. Will include metrics tracked (and how to filter through noise), when and how to migrate, operational optimizations such as upgrades that can be incorporated, and how to deal with server technology that evolves faster than server lifetimes
MySQL replication has been a critical part of scaling Facebook's storage infrastructure. However, it brings with it the fear of divergent replicas. This session discusses a new tool that detects divergences, identifies inconsistent data, and helps repairs defects.
The Facebook database engineering team works with the community and on its own to make MySQL better for data center deployments. This work is visible in the Facebook patch, bugs fixed in official MySQL and features sponsored in other distributions. We will describe work to support a large number of large databases. We focus on backup, replication and quality of service.
A real-world example of how re-sharding and table partitioning cut load data times in Facebook's analytics infrastructure from greater than 24 hours to less than 5 minutes.
Getting sharding right is crucial for achieving high scale with MySQL on commodity hardware like we do at Facebook. We will overview sharding best practices, and show some examples of both successful and unsuccessful methods at sharding MySQL.