1~2 minutes downtime - notified in advance - need to notify customers - customers often want to reschedule (possible within 28 days of date) - high support overhead ~5 min/year downtime (0.999%) is itself not a big issue, but highly opinionated customers are.
from hitting a limit that scaling up cannot solve - But we have had on-premise customers who are looking for very high scalability High availability - Ability to distribute cross-region (and globally) opens up new options for us - Being able to scale up/out the DB without downtime is an advantage
We have discussed Vitess, Spanner, CochroachDB, YugabyteDB, TiDB since 2020 - Prefer managed solutions (don’t want to hire DBA team) - Need good compatibility with our Java DB middleware (Mysql/Postgres) - Ideally should offer on-premise DB solution for our on-premise customers - Good support and partnership opportunities
further: - Enabled bunch of flags (for the most part, I didn’t know what these flags did) - We tested upgrading the Yugabyte Managed cluster while running load test - We achieved satisfactory results
unsupported data type change - At that time, PostgreSQL supported this one, not Yugabyte - Yugabyte team provided support and helped us to fix our issue in recent version
are planning on running a pilot project with customer cooperation - Evaluation of real-world use of Yugabyte-backed Authlete SaaS Yugabyte officially supported for on-premise customers - We are communicating to our customers that Yugabyte is a supported DB option General availability as an option for SaaS customers - Upon completion of pilot project, we are expecting to offer Yugabyte as an HA option