Key Concepts

MemStore is a managed cache database service that allows you to easily utilize Redis OSS without the need for separate installation processes. The key concepts of the MemStore service are as follows:

Key concept	Description
Cache Only	A cache service that stores all data in memory without using disk storage
Cluster	A group of multiple shards, each consisting of one Primary node and multiple Replica nodes - You can choose to enable or disable cluster mode based on your use case
Shard	A unit that partitions the dataset within a cluster, composed of multiple nodes. Each node plays the role of either a Primary or Replica. - Add shard: Add a shard when the selected cluster experiences increased load - Delete shard: Remove unnecessary shards from the selected cluster
Node	A virtual machine (VM) with Redis OSS installed that serves as a Primary or Replica - Node types: Select a suitable node type based on data size and load when creating a cluster - Node lifecycle and status: Understand node conditions based on status - Add node: Add a Replica node to reduce read load in a selected cluster - Delete node: Remove unnecessary nodes from a selected cluster
High availability	Ensures service continuity by automatically responding to node failures - If a Primary node fails, the Replica with the most up-to-date data is automatically promoted to Primary (Automatic Failover), and a new Replica is created to maintain the desired Replica count (Auto Healing) - If a Replica node fails, it is automatically replaced with a new node
Promotion	When high availability is enabled: - Promotion is automatic and cannot be requested manually When high availability is disabled: - Replica nodes can be manually promoted to Primary
Read/write-only endpoint provision	When cluster mode is disabled: Provides a Primary endpoint for read/write access and a read-only endpoint When cluster mode is enabled: Does not provide a read-only endpoint; instead, each node provides an individual endpoint

Cache Only

Redis in MemStore is used as a cache only, storing all data in memory and not on disk.

Cluster

A cluster is a group of one or more nodes that are automatically managed. A cluster consists of multiple shards and nodes, with each node functioning as either a Primary or a Replica. Shards and Replica nodes can be added or deleted as needed.

Cluster lifecycle and status values

When using cluster mode

info

When using cluster mode, shard and node states are dependent on the cluster's state and share the same status.

Lifecycle diagram for cluster/shard/node in cluster mode

Status	Description
Init	Initial state before cluster creation begins
Creating	Resources are being created
Running	Cluster is running normally
Modifying	Cluster configuration is being changed (e.g., adding/removing shards or nodes)
Deleting	Cluster is being deleted
Failed	A permanent failure has occurred. Cluster cannot be recovered and can only be deleted.

When not using cluster mode

Lifecycle diagram for non-cluster mode

Status	Description
Init	Initial state immediately after cluster creation
Provisioning	Nodes are preparing required resources
Replicating	Node roles are being configured and endpoints are being generated
Running	Service is fully operational and accessible
ScalingIn	Replica node is being deleted
SwitchingOver	A Replica node is being promoted to Primary
ScalingOut	New nodes are being added as Replicas
Recovering	Auto-recovery actions are being performed for a failed node in an HA cluster
Modifying	Security group change is in progress
Terminating	Cluster deletion is in progress
Terminated	Cluster has been deleted
Warning	Temporary issue occurred; will return to normal once resolved
Failed	A permanent failure occurred. Cluster cannot be recovered and must be deleted.

Shard

A shard is a group of nodes in cluster mode that stores a partition of the overall dataset.
Clusters can have 1 to 12 shards. Each shard must have between 2 and 5 nodes when using cluster mode.

Node

A node is a VM that belongs to a cluster. When created or deleted, nodes automatically execute required operations and update their status. Each node is assigned a role of either Primary or Replica. When cluster mode is disabled, a maximum of 6 nodes can be added.

Role	Description
Primary	Read/write node used for client operations
Replica	Read-only node used to offload traffic from the Primary

Node types

Choose a node type based on data volume and load. More specialized types may be supported in the future.

Node type	vCPU	Memory (GB)
m2a.large	2	8
m2a.xlarge	4	16
m2a.2xlarge	8	32
m2a.4xlarge	16	64
m2a.8xlarge	32	128
m2a.12xlarge	48	192
m2a.16xlarge	64	256
m2a.24xlarge	96	384
r2a.large	2	16
r2a.xlarge	4	32
r2a.2xlarge	8	64
r2a.4xlarge	16	128
r2a.8xlarge	32	256
r2a.12xlarge	48	384
r2a.16xlarge	64	512
r2a.24xlarge	96	768
t1i.small	2	2
t1i.medium	2	4
t1i.large	2	8
t1i.xlarge	4	16
t1i.2xlarge	8	32

info

t1i instances may take several minutes to transition to Running after initial creation.
Also, if CPU credits are insufficient, workloads may not function properly.

Node lifecycle and status values

• When using cluster mode:

  • Node lifecycle follows the cluster status. See Cluster lifecycle and status.

• When not using cluster mode:

Node lifecycle diagram

Status	Description
Init	Node creation process starts
Provisioning	Infrastructure provisioning is in progress
Starting	Service installation and startup preparation
Running	Node is running and accessible
Modifying	Security group modification is in progress
Terminating	Node is being deleted upon request
Terminated	Node has been deleted and is no longer accessible
Warning	Temporary issue occurred; will recover automatically
Failed	A permanent failure occurred; node must be deleted

High availability

MemStore supports high availability (HA) for improved reliability and seamless service. When enabled, HA automatically recovers from node failures.

• When using cluster mode: HA is always enabled and cannot be disabled.
• When not using cluster mode: You can enable or disable HA at cluster creation or during operations.
• HA-enabled clusters automatically include newly added nodes in the HA configuration.

Node failures

Case	Description
Primary node failure	The most up-to-date Replica is automatically promoted to Primary (Automatic Failover), and a new Replica is created (Auto Healing).
Replica node failure	The failed Replica is automatically replaced to maintain Replica count.

Promote node

• When using cluster mode or HA: Promotion is handled automatically and cannot be triggered manually.
• When HA is disabled: You can manually promote a Replica to become the Primary if needed.

Read/write-only endpoint provision

When a cluster is created, it provides a read-only endpoint for client access.
This endpoint is accessible from the selected private subnet and comes with secure access control and firewall rules.

info

When using cluster mode, read-only endpoints are not provided. Use each node's endpoint instead.

Backup

You can back up a cluster and restore it when needed.
Currently, backup is supported only when cluster mode is disabled.

Backup status values

Status	Description
Available	Backup is complete and ready for restore, copy, or export
Pending	Backup request has been accepted but not yet started
Creating	Backup is in progress or being copied from another backup
Restoring	Backup is being used for provisioning
Copying	Backup is being copied to create another backup
Exporting	Backup is being exported to the user’s Object Storage
Error	An error occurred during the backup process
Deleting	Backup is being deleted

Parameter application status

Status	Description
Pending	Waiting to apply parameters
Applying	Parameters are being applied to the cluster
In-Sync	Parameters have been successfully applied
Error-Sync	Parameter application failed. You can retry, which will change the status to Applying.
Suspended	Waiting due to a failure in a previous sequential application. You can retry, which will change the status to Applying.