Overview

Provisioning

replication-manager-pro can provision and manage the complete lifecycle of database and proxy infrastructure — creating service definitions, deploying containerised or bare-metal instances, bootstrapping replication topology, and tearing everything down — across five orchestration backends.

Requires: replication-manager-pro flavor. The osc release supports onpremise and local orchestrators only.

Orchestration Backends

Orchestrator prov-orchestrator Description
OpenSVC opensvc Service cluster with Docker/Podman micro-services across a pool of agents. Primary supported backend.
Kubernetes kube Native K8s deployment using a kubeconfig file.
SlapOS slapos Distributed service mesh from Nexedi.
On-Premise SSH onpremise SSH-based bootstrap onto existing bare-metal or VM hosts. No agent required.
Local local Processes started directly on the replication-manager host. For testing only.

What Can Be Provisioned

Database Services

  • Containers: MariaDB, MySQL, Percona — any Docker/Podman image
  • Resources: CPU cores, memory, shared memory, tmpfs
  • Disk: size, filesystem (ext4, ZFS, Btrfs), pool, compression, device type
  • Snapshots: daily snapshots on preferred primary, configurable retention
  • Network: interface, netmask, gateway, CNI virtual networks
  • Bootstrap: replication topology, initial data load (SQL/CSV), benchmark table setup
  • Config: dynamic config application, binary log settings, max connections, expire-logs-days

Proxy Services

All proxy types share resource (memory, CPU, disk) and network configuration:

Proxy Notes
HAProxy TCP load balancer, read/write split
ProxySQL Query routing, connection pooling
MaxScale MariaDB-native proxy with binlog routing
ShardProxy Spider-based sharding proxy
MySQL Router Lightweight routing for InnoDB Cluster
Sphinx Full-text search engine as a service

Service Plans

Pre-defined plans set the number of database nodes and resource tiers in a single setting (prov-service-plan). Plans are registered in a central registry and can be switched dynamically, triggering automatic cluster resize.

Bootstrap Workflow

When you provision a cluster, replication-manager executes in order:

  1. Provision all database service containers/processes
  2. Provision all proxy service containers/processes
  3. Bootstrap the replication topology
  4. Create the sponsor/monitoring user if configured
  5. Wait for proxies to synchronise with the new topology
  6. Optionally initialise a benchmark (Sysbench) table

Serialised provisioning mode (prov-db-service-plan-batch) controls whether services are started one-by-one (dependency ordering) or in parallel.

Configuration Guide

See the sub-sections for detailed configuration of the Software Configurator, Orchestrators, and Service Plans.