Bryan Beaudreault <bbeaudreault@...>
I'm an engineer at HubSpot who was part of the decision to adopt vitess over a year ago and have contributed to the project both internally and externally. In that time we have committed numerous features and improvements, and moved along a careful process of evaluation followed by bringing Vitess to production. As Sugu said, we have multiple full time employees working on Vitess, all of which have committed code to Vitess.
In terms of our architecture, we operate microservices at HubSpot, and follow a typical pattern of different microservices getting their own database. We have over 200 separate database clusters of various sizes, each with a QA and production footprint. For the platform teams at HubSpot QA is effectively production, as we have a separate test environment where we stage changes (in total: staging, qa, then prod). Understandably, the definition of production usage should be based on actual customer usage. We have had nearly 100% of QA migrated for months now, and started the prod migration in the last month or so. At this time we have 5 production database clusters migrated to Vitess. That is a small percentage of the total, but these are real applications, powering services used by real HubSpot customers, and we treat prod as mission critical.
We will be continuing the prod rollout for a few months, but we are well past the evaluation phase. At this point we are fully committed to vitess, with production users, support staff, and active development.
Hope that helps.