Amazon OpenSearch Service introduces capacity optimized blue/green deployments
Amazon OpenSearch Service now offers a Capacity Optimized option for blue/green deployments, ensuring domain updates can complete even when available instance capacity is less than required. Updates are performed in incremental batches, reducing the number of additional instances needed during the process.
Amazon OpenSearch Service uses a blue/green deployment process when updating domains — creating an idle copy of the original environment, applying updates, and routing traffic to the new environment once complete. This minimizes downtime and preserves the original environment as a fallback. Until now, blue/green deployments required 100% instance capacity upfront. For example, for a cluster with 100 data nodes, another 100 nodes were needed to proceed. If sufficient capacity was unavailable, customers had to wait and retry later.
Now, customers can choose between two deployment strategies. The default Full Swap option maintains current behavior, requiring full capacity upfront for the fastest deployment. The new Capacity Optimized option attempts a full capacity deployment first, but automatically falls back to batch deployment if capacity is insufficient. OpenSearch Service determines the appropriate batch size based on cluster size and available instances. Because updates are applied in batches, this option may take longer than a full-swap deployment. It is recommended for large clusters with 30 or more data nodes. Customers can select their preferred option in the deployment configuration settings via the OpenSearch Service console or API.
The Capacity Optimized option is available for all OpenSearch and Elasticsearch versions, in all Amazon Web Services regions where Amazon OpenSearch service is available, including the Amazon Web Services China (Beijing) Region, operated by Sinnet and the Amazon Web Services China (Ningxia) Region, operated by NWCD. Please refer to the Region Table for more information about Amazon OpenSearch Service. To learn more, visit the documentation page.