OceanBase Glossary

Concise definitions for OceanBase architecture and operations. Each entry links to official documentation and related OceanDB Pro guides where available.

OceanBase Database

A distributed relational database that stores data in log streams replicated with Paxos. It supports MySQL and Oracle compatibility modes.

Why it matters: Understanding the product boundary helps you choose compatibility mode, sizing, and migration tooling.

Tenant

An isolated database instance inside an OceanBase cluster. Each tenant has its own resource units, compatibility mode, and data namespace.

Why it matters: Multi-tenant clusters rely on correct tenant boundaries for isolation, quotas, and blast-radius control.

Zone

A failure domain, typically mapped to a data center or availability zone, that hosts observer replicas for Paxos quorums.

Why it matters: Zone layout determines whether the cluster survives single-site or single-region failures.

Region

A geographic grouping of zones used for disaster recovery and multi-region quorum design.

Why it matters: Cross-region replication affects latency, RPO, and failover behavior.

Unit

A resource slice (CPU, memory, disk) allocated to a tenant on a specific observer node.

Why it matters: Unit placement drives performance locality and failure isolation per tenant.

Unit config

A template defining CPU, memory, and storage limits for units assigned to a resource pool.

Why it matters: Incorrect unit configs cause OOM, throttling, or wasted hardware.

Resource pool

A pool of units with the same unit config distributed across zones for a tenant.

Why it matters: Pools define how tenant capacity scales and spreads across the cluster.

Locality

Rules that specify how many replicas of a log stream or partition exist in each zone.

Why it matters: Locality drives quorum math and read locality for HTAP workloads.

Tablet

A partition of user data mapped to a log stream. Tablets are the unit of data distribution inside a tenant.

Why it matters: Hot tablets affect single log stream leaders and migration parallelism.

Log stream

An independently replicated shard that holds a portion of tablet data and its own Paxos log.

Why it matters: Leader election, replication lag, and failover all operate per log stream.

Leader

The replica in a log stream that accepts writes and coordinates Paxos commits for that stream.

Why it matters: Leader loss triggers election; misconfigured quorums can block writes.

Follower

A non-leader replica that receives and replays log entries from the leader for a log stream.

Why it matters: Follower lag and role transitions indicate replication or network problems.

Paxos

The consensus protocol OceanBase uses to replicate log streams and elect leaders among replicas.

Why it matters: Quorum size and zone layout determine fault tolerance and failover behavior.

MemStore

In-memory storage for recent writes in an observer before data is flushed to stable storage.

Why it matters: MemStore pressure causes write stalls and triggers minor freezes.

Minor freeze

A process that flushes memstore data to disk as a minor SSTable when memory thresholds are reached.

Why it matters: Frequent minor freezes can signal memstore limits or hot write patterns.

Minor compaction

Compaction that merges minor SSTables to reduce read amplification before major compaction.

Why it matters: Compaction backlog increases read latency and disk usage.

Major compaction

Full compaction that rewrites SSTables into a stable baseline, often scheduled during maintenance windows.

Why it matters: Major compaction affects I/O and should be planned for large tenants.

OCP (OceanBase Cloud Platform)

Management software for deploying, monitoring, and operating OceanBase clusters.

Why it matters: Many production teams use OCP for failover visibility, backups, and tenant operations.

OBD (OceanBase Deployer)

Command-line tooling to deploy and bootstrap OceanBase clusters, often used before or alongside OCP.

Why it matters: Bootstrap and topology mistakes at deploy time are expensive to fix later.

OMS (OceanBase Migration Service)

Data migration and synchronization service for moving data from supported sources into OceanBase.

Why it matters: OMS capabilities define feasible migration paths and cutover strategies.

ODP (OceanBase Database Proxy)

Connection proxy (OBProxy) that routes client SQL to the correct observer and log stream leader.

Why it matters: Applications should typically connect through the proxy for leader failover tolerance.

MySQL mode

OceanBase compatibility mode that accepts a large subset of MySQL SQL and wire protocol behavior.

Why it matters: Migration from MySQL depends on dialect coverage and driver settings in this mode.

Oracle mode

OceanBase compatibility mode targeting a subset of Oracle SQL and PL/SQL features.

Why it matters: Oracle migrations require explicit assessment of unsupported constructs.

HTAP

Hybrid transactional and analytical processing: running OLTP and analytical queries on shared or coordinated storage.

Why it matters: HTAP shapes resource isolation, column store use, and workload scheduling.

RootService

Cluster metadata service that coordinates tenant creation, partition balance, and global cluster state.

Why it matters: RootService health is required for topology changes and some recovery operations.