What every engine gets
- Plans with risk and disruption for every change, approvals by policy and resumable operations.
- Encrypted backups on by default, restores to a new cluster and scheduled restore drills.
- Built-in metrics, a dashboard, alert rules with runbooks and a write canary.
- TLS required by default, with a certificate authority per cluster.
- Endpoints through HAProxy on every member, with an optional virtual IP (MongoDB drivers connect directly).
- Memory and connection budgets checked before a plan runs, and a data identity guard on every member.
Never inventing consensus
Patroni with etcd, MongoDB replica-set elections, Sentinel, Galera and Garage already solve leader election and
replication, and have been tested by their communities for years. ElastiCluster orchestrates them: it installs
them as containers pinned by digest, configures them consistently, watches them and changes them through plans.
The control plane is never in the data path, so the native tools (patronictl, mongosh,
valkey-cli) keep working and tell you the truth.
Which engine for what
- PostgreSQL for relational workloads, the flagship and first engine.
- MongoDB for document data in replica sets.
- Valkey and Dragonfly for caches, sessions and queues.
- MariaDB Galera for MySQL-compatible applications.
- Garage for S3-compatible object storage and backup targets.
Later
Valkey cluster mode and MongoDB sharding are planned for v1.1. Candidates after v1.0 include SeaweedFS,
ClickHouse and FerretDB. Vector search is covered by pgvector in PostgreSQL.