01
- Rescue
- A cluster that will not elect a master, tservers that keep crashing, or a split loop that is dragging latency up. We settle the quorum first and dig into the cause after.
02
- Schema and workload review
- Tables reviewed for hot partitioning, primary keys examined for even distribution, and indexes checked against what queries actually run.
03
- Multi-region topology design
- Placement policies, preferred leaders, follower reads and read replicas arranged so latency meets the SLA without giving up durability.
04
- Migration from Postgres
- ysql_dump, logical replication or a blue/green cutover depending on the size and uptime budget, with the application driver and pool settings adjusted for a distributed backend.
05
- Upgrades and backups
- Rolling upgrades rehearsed on a copy of the cluster, distributed snapshots and PITR configured, and restore drills run before anyone needs them.
06
- Managed Tech
- Daily health reviews, tablet balance kept in check, upgrade planning and an engineer to call when a Raft log starts misbehaving in production.