Cache

The Cache service provides managed Redis instances with support for standalone, sentinel (HA), and cluster (sharded) modes. The Cache Agent handles installation, configuration, replication, and automatic failover.

Modes

## Features

• Persistence — configurable AOF (Append Only File) and RDB (snapshot) persistence
• Eviction policies — choose from noeviction, allkeys-lru, volatile-lru, allkeys-random, and more
• Replication monitoring — the agent tracks replication lag and replica health
• Auto-failover — in sentinel mode, the agent detects primary failure and coordinates promotion

Create a Redis Instance

Standalone

agentmetal redis create --name dev-cache --mode standalone

Sentinel (High Availability)

agentmetal redis create \
  --name cache \
  --mode sentinel \
  --nodes 3

This creates one primary and two replicas, with three sentinel processes monitoring the group.

Cluster (Sharded)

agentmetal redis create \
  --name session-store \
  --mode cluster \
  --nodes 6

Creates a 6-node Redis Cluster with 3 masters and 3 replicas, distributing hash slots across masters.

Configuration

Set Redis configuration parameters:

agentmetal redis config set cache \
  --param maxmemory=2gb \
  --param maxmemory-policy=allkeys-lru \
  --param appendonly=yes

Get Connection Info

agentmetal redis get cache --connection-string

For sentinel mode, the output includes the sentinel addresses and master name for client configuration.

What the Agent Manages

The Cache Agent performs these operations:

1. Redis installation — installs Redis on provisioned VMs and applies OS-level tuning (vm.overcommit_memory, transparent huge pages)
2. Sentinel configuration — deploys Redis Sentinel processes, configures quorum, and sets up monitoring of the primary
3. Cluster formation — initializes Redis Cluster, assigns hash slots, and adds replica nodes
4. Failover — detects primary failure via sentinel, coordinates promotion, and updates connection information
5. Persistence management — configures AOF/RDB schedules based on your requirements and available disk
6. Health monitoring — checks memory usage, connected clients, replication lag, and keyspace hit rate

Scaling

Scale a cluster by adding more shards:

agentmetal redis scale session-store --nodes 9

The agent adds new nodes, rebalances hash slots, and ensures data migration completes before marking the operation as done.