November 2024 Summaries

In this conclusion of a three-part blog series on throttlers, we discuss client identities, cooperation, prioritization, and constraints issues. Clients should identify themselves to the throttler for analysis or investigation purposes, as well as operational purposes such as prioritizing one job over others or tuning down specific clients. We propose a safer way of playing within the rules by considering prioritization or de-prioritization through a dice roll mechanism. Throttling on different metrics can be beneficial but should be done with practical considerations in mind. Exemption scenarios may make sense in certain situations, and categorization and breakdown of client identities can help in analyzing throttler access patterns by categories of requests. Lastly, we emphasize the importance of dynamic control of the throttler and the ability to prioritize or push back some requests or jobs in production systems.

PlanetScale has introduced new workflows functionality that enables users to perform actions on their databases through predefined steps. The first workflow allows horizontal scaling of databases by moving tables to a sharded keyspace, all within the PlanetScale dashboard. This feature is similar to MoveTables in Vitess and is now available for all PlanetScale plans. Sharding is a proven architecture used by large web properties to scale their databases, and PlanetScale aims to be the most reliable and scalable platform for running relational databases. The new workflows functionality enables users to easily migrate existing data to sharded keyspaces and switch production traffic between them with no downtime.