Liferay Enterprise Search Cross-Cluster Replication Released

Liferay is pleased to announce the release of the Cross-Cluster Replication (CCR) module for Liferay Enterprise Search subscribers. In short, the CCR module replicates Liferay DXP's search engine cluster across data centers.

This is a feature request that Liferay often hears from customers, but was not previously possible because unlike Liferay DXP clusters, all the nodes in an Elasticsearch clusters must be on the same local network. Elastic has since released their Cross-Cluster Replication feature to natively replicate data between clusters, and Liferay's CCR module leverages this feature to provide high availability, disaster recovery, and geoproximity performance optimization.

We believe this new feature will be particularly valuable for mission-critical Liferay DXP deployments where search capabilities require maximum resilience and uptime. For example, a business using Liferay Commerce can drastically minimize the unavailability of searching for products, while improving response times by co-locating their Liferay DXP and Elasticsearch cluster deployment to their customers' physical geography.

How Does It Work?

Liferay's CCR module leverages Elasticsearch's Cross-Cluster Replication feature, which uses an active-passive model to replicate search indices (the leader indices) from one Elasticsearch cluster to other clusters (the follower indices). Liferay DXP nodes can be configured to read from any of the follower indices, while all writes are sent to the leader indices.


The diagram above shows a sample deployment leveraging CCR. The US data center contains one Liferay DXP node and the leader indices, whereas the Hungary data center contains two Liferay DXP nodes and the follower indices. All clustered Liferay DXP nodes write to the leader indices in the US, while nodes DXP2 and DXP3 read from the follower indices in Hungary.

If the Elasticsearch cluster in Hungary experiences downtime, node DXP1 will continue to be fully operational. Reads for DXP2 and DXP3 will fail, but writes will continue to work. These two nodes can be reconfigured to read from the US Elasticsearch cluster until the Hungary cluster is back online, at which point all missing writes from the leader indices will replicate to the follower indices.

Alternatively, if the Elasticsearch cluster in the US experiences downtime, reads will continue to work for nodes DXP2 and DXP3, but reads will fail for DXP1, and writes will fail for all nodes. DXP1 can be reconfigured to read from the Hungary Elasticsearch cluster until the US cluster is back online, at which point a reindex will be required to populate the missing writes.

Additionally, the deployment above reduces the latency of search requests as the DXP node and Elasticsearch are physically closer together. 

How Do I Get It?

The Cross-Cluster Replication module is available to Liferay Enterprise Search subscribers via the Help Center. Downloads for Liferay DXP 7.3 and 7.2 are available now, and a module for Liferay DXP 7.1 is coming soon. Documentation is available here.

We're excited for our Liferay Enterprise Search subscribers to try out this new feature! If you are interested but are not a Liferay Enterprise Search subscriber, please contact your sales representative.

0