Skip to content

Deployment and Operations Backup and Recovery

Jump to bottom
github-actions[bot] edited this page Aug 3, 2026 · 3 revisions

Backup and Recovery

Referenced Files in This Document

Table of Contents

  1. Introduction
  2. Project Structure
  3. Core Components
  4. Architecture Overview
  5. Detailed Component Analysis
  6. Dependency Analysis
  7. Performance Considerations
  8. Troubleshooting Guide
  9. Conclusion
  10. Appendices

Introduction

This document provides comprehensive backup and recovery guidance for Kairos MCP, covering:

  • Data export/import procedures for memory store, artifacts, and workflow states
  • Snapshot creation and restoration for the Qdrant vector database
  • PostgreSQL backup strategies
  • Redis cache considerations
  • Artifact storage backup
  • Disaster recovery procedures
  • Data migration strategies
  • Testing recovery processes
  • Automated backup scheduling and retention policies

The goal is to enable operators to protect data integrity, minimize downtime, and reliably restore services across environments.

Project Structure

Kairos MCP integrates multiple persistent components:

  • Memory store (vector search via Qdrant)
  • Relational data via PostgreSQL (managed by an operator)
  • Caching and pub/sub via Redis (managed by an operator)
  • Artifacts stored on disk or external storage
  • HTTP APIs and CLI tools for exporting and dumping data
graph TB
subgraph "Kairos MCP"
API["HTTP API"]
Tools["CLI / Tools"]
Mem["Memory Store"]
Qdrant["Qdrant Service"]
Redis["Redis Cache"]
PG["PostgreSQL"]
Artifacts["Artifact Storage"]
end
API --> Mem
Tools --> Mem
Mem --> Qdrant
API --> Redis
API --> PG
API --> Artifacts
Loading

[No sources needed since this diagram shows conceptual workflow, not actual code structure]

Core Components

  • Qdrant snapshot management for vector index snapshots
  • Dump and export utilities for memory store content and artifacts
  • Redis configuration for caching and pub/sub behavior
  • Helm templates for PostgreSQL and Redis deployment configurations

Key responsibilities:

  • Create and restore Qdrant snapshots
  • Export and import memory store data and artifacts
  • Configure and back up PostgreSQL and Redis
  • Provide operational scripts for testing and seeding

Section sources

Architecture Overview

Backup and recovery spans several subsystems:

  • Vector index snapshots are managed through dedicated snapshot endpoints and service logic
  • Memory store data can be dumped and exported via HTTP and CLI interfaces
  • Artifacts are downloadable through specific routes
  • PostgreSQL and Redis are provisioned via Kubernetes operators with their own backup mechanisms
sequenceDiagram
participant Admin as "Operator"
participant API as "HTTP API"
participant SnapSvc as "Snapshot Service"
participant Qdrant as "Qdrant"
participant Export as "Export/Dump"
participant Artifacts as "Artifacts"
participant PG as "PostgreSQL"
participant Redis as "Redis"
Admin->>API : "Create snapshot"
API->>SnapSvc : "Initiate snapshot"
SnapSvc->>Qdrant : "Create snapshot"
Qdrant-->>SnapSvc : "Snapshot metadata"
SnapSvc-->>API : "Snapshot ID"
API-->>Admin : "Snapshot created"
Admin->>API : "Restore snapshot"
API->>SnapSvc : "Restore from snapshot"
SnapSvc->>Qdrant : "Restore snapshot"
Qdrant-->>SnapSvc : "Restoration status"
SnapSvc-->>API : "Restoration complete"
API-->>Admin : "Restored"
Admin->>API : "Dump memory store"
API->>Export : "Generate dump"
Export-->>API : "Dump payload"
API-->>Admin : "Download dump"
Admin->>API : "Export artifacts"
API->>Artifacts : "List and package"
Artifacts-->>API : "Archive"
API-->>Admin : "Download archive"
Note over PG,Redis : "Back up using operator-native mechanisms"
Loading

Diagram sources

Detailed Component Analysis

Qdrant Snapshot Management

  • Creation: The snapshot service coordinates with Qdrant to create a consistent snapshot of the vector index.
  • Restoration: The service triggers a restore operation from a specified snapshot and reports completion.
  • API exposure: HTTP endpoints wrap these operations for operator interaction.
classDiagram
class SnapshotService {
+createSnapshot() string
+restoreSnapshot(snapshotId) void
}
class HttpSnapshotApi {
+postCreateSnapshot() Response
+postRestoreSnapshot() Response
}
SnapshotService <.. HttpSnapshotApi : "invoked by"
Loading

Diagram sources

Section sources

Memory Store Dump and Export

  • Dump: Generates a structured representation of memory store contents for archival or migration.
  • Export: Produces artifact bundles and related metadata for portability.
  • HTTP endpoints expose both dump and export capabilities for automation.
flowchart TD
Start(["Start"]) --> Choose["Choose Operation"]
Choose --> |Dump| DumpFlow["Call Dump Endpoint"]
Choose --> |Export| ExportFlow["Call Export Endpoint"]
DumpFlow --> GenerateDump["Generate Dump Payload"]
ExportFlow --> PackageArtifacts["Package Artifacts and Metadata"]
GenerateDump --> ReturnDump["Return Downloadable Dump"]
PackageArtifacts --> ReturnExport["Return Downloadable Archive"]
ReturnDump --> End(["End"])
ReturnExport --> End
Loading

Diagram sources

Section sources

Redis Cache Considerations

  • Role: Provides caching and pub/sub functionality; typically ephemeral but may hold transient state.
  • Configuration: Connection settings and behavior are defined in service modules.
  • Backup strategy: Since caches are often transient, focus on replication and failover rather than persistent backups.
graph LR
App["Kairos MCP"] --> Cache["Redis Cache"]
App --> PubSub["Redis Pub/Sub"]
Loading

Diagram sources

Section sources

PostgreSQL Backup Strategies

  • Provisioning: Managed via a PostgreSQL cluster custom resource.
  • Operator-backed backups: Use the operator’s built-in backup and restore features to capture consistent snapshots of relational data.
  • Retention: Configure retention policies at the operator level to manage storage costs and compliance.
graph TB
App["Kairos MCP"] --> PGCluster["PostgreSQL Cluster CR"]
PGCluster --> Backup["Operator Backup Job"]
Backup --> Storage["Object Storage"]
Loading

Diagram sources

Section sources

Artifact Storage Backup

  • Artifacts are accessible via download routes and can be packaged for export.
  • Ensure underlying storage (filesystem or object store) is backed up independently to preserve binary assets.
graph TB
App["Kairos MCP"] --> Routes["Export Download Routes"]
Routes --> Artifacts["Artifact Storage"]
Artifacts --> Backup["Storage Backup"]
Loading

Diagram sources

Section sources

Dependency Analysis

  • HTTP API depends on snapshot service, dump/export tools, and artifact routes.
  • Memory store interacts with Qdrant for vector operations and may use Redis for caching.
  • PostgreSQL and Redis are provisioned via Kubernetes operators and should be backed up using operator-native mechanisms.
graph TB
API["HTTP API"] --> Snap["Snapshot Service"]
API --> Dump["Dump Tool"]
API --> Export["Export Tool"]
API --> ArtRoutes["Export Download Routes"]
API --> Redis["Redis"]
API --> PG["PostgreSQL"]
Snap --> Qdrant["Qdrant"]
Loading

Diagram sources

Section sources

Performance Considerations

  • Scheduling snapshots during low-traffic windows to reduce contention.
  • Staggering exports and dumps to avoid I/O spikes.
  • Using incremental or differential backups where supported by operators.
  • Monitoring snapshot size and restore time to plan capacity and RTO/RPO targets.

[No sources needed since this section provides general guidance]

Troubleshooting Guide

  • Verify snapshot availability and integrity before attempting restore.
  • Validate that dump payloads match expected schema when importing into target environments.
  • Confirm Redis connectivity and TTL settings if restoring cached state is required.
  • Check PostgreSQL operator logs for backup job success and retention policy enforcement.

Section sources

Conclusion

A robust backup and recovery strategy for Kairos MCP combines:

  • Consistent Qdrant snapshots for vector indexes
  • Structured dumps and exports for memory store and artifacts
  • Operator-managed backups for PostgreSQL and Redis
  • Clear disaster recovery procedures and tested restoration workflows
  • Automated scheduling and retention aligned with organizational policies

[No sources needed since this section summarizes without analyzing specific files]

Appendices

Operational Procedures

Create Qdrant Snapshot

  • Trigger snapshot creation via the snapshot endpoint.
  • Record the returned snapshot identifier for later restoration.

Section sources

Restore Qdrant Snapshot

  • Invoke the restore endpoint with the snapshot identifier.
  • Monitor restoration status and verify vector search consistency post-restore.

Section sources

Export Memory Store and Artifacts

  • Use the dump endpoint to generate a structured dump of memory store data.
  • Use the export endpoint to package artifacts and associated metadata.
  • Download resulting payloads for archival or migration.

Section sources

Import Test Snapshot

  • Utilize provided scripts to seed or import test snapshots for validation.

Section sources

PostgreSQL Backup and Restore

  • Configure backups using the PostgreSQL operator’s custom resource definitions.
  • Schedule periodic backups and define retention policies at the operator level.
  • Restore from operator-managed backups following standard operator procedures.

Section sources

Redis Failover and Resilience

  • Deploy Redis with failover support via the Redis operator.
  • Focus on high availability and replication rather than persistent backups for ephemeral cache data.

Section sources

Automated Backup Scheduling and Retention Policies

  • Schedule Qdrant snapshots periodically using external orchestrators (e.g., CronJob).
  • Configure PostgreSQL operator backup schedules and retention windows.
  • Periodically trigger memory store dumps and artifact exports based on change frequency.
  • Define retention policies for all backups to balance cost and compliance requirements.

[No sources needed since this section provides general guidance]

KAIROS MCP

Clone this wiki locally