Shard Balancing: Moving Shops Confidently with Zero-Downtime at Terabyte-scale

Shard Balancing: Moving Shops Confidently with Zero-Downtime at Terabyte-scale

摘要

Shopify’s infrastructure supports over 1.7 million merchants during their entrepreneurship journey. A key component of the current infrastructure is the underlying fleet of MySQL database shards that together persist every shops’ critical data. As traffic patterns change and new merchants onboard onto the platform, it’s possible that resource intensive shops end up living in the same shards. Certain database shards become unbalanced in their database utilization, shop traffic, and load. It’s important to ensure the shards remain well-balanced to mitigate risk of database failure, improve productivity of the wider infrastructure, and ultimately guarantee buyers can always access their favorite shops. This post explains how we’re able to balance our MySQL shards by migrating shops across shards—entirely online and with virtually zero consumer-facing downtime.