Content Store Replication

From AlfrescoWiki

Contents 1 Introduction 2 Existing Components 2.1 ReplicatingContentStore 2.2 ContentStoreReplicator 3 Required Components 3.1 ContentStore Replicating over HTTP 3.1.1 Background 3.1.2 Requirements 3.1.3 Implementation 3.1.4 Issues

[edit] Introduction

Clustered servers have to share content across the instances by replicating the content. This article discusses some of the available options, as well as the design of future or desired components.

[edit] Existing Components

[edit] ReplicatingContentStore

[edit] ContentStoreReplicator

[edit] Required Components

[edit] ContentStore Replicating over HTTP

[edit] Background

In some cluster configurations, the option to share content directly via a filesystem(s) doesn't exist. If the machines are all live within the cluster, then it is not possible to push the content from one server to another using rsynch or some other delayed mechanism. Instead, the content must be pulled onto the server that requires it. Alfresco servers already have a DownloadContentServlet that is able to access content using NodeRef and path based references.

[edit] Requirements

1. The ability to remotely request content from a server using a ContentStore content URL of the form store://....
2. The notion of a read-only ContentStore

[edit] Implementation

1. Additional get method on ContentService to access node-independent stream. Access must be limited to admin privileged user only.
2. Extend ContentDownloadServlet to access content via new new URL.
3. Add ReadOnlyContentStore flag interface and ensure that it gets respected by ReplicatingContentStore.
4. Cluster testing
5. Estimated (without significant issues): 2 days implementation, 2 days testing

[edit] Issues

1. Sharing of authentication tickets or force logins to each machine providing content in the cluster.
2. Configuration of machines available in the cluster.