Opportunity Name:
Convert Elasticsearch 7.9+ Instances to Graviton
AWS Resource Type:
Amazon OpenSearch Service
Opportunity Description:
This opportunity upgrades OpenSearch domains with Elasticsearch version 7.9 and above to Graviton2. Graviton2-based instances are 10% cheaper than previous generation Intel-based instances (i.e., M5, C5, R5) while providing a 38% improvement in indexing throughput, a 50% reduction in indexing latency, and a 30% improvement in query performance.
Criteria for identifying the opportunity:
This Finder identifies OpenSearch instance opportunities if all of the following criteria are met:
- instance is running on-demand (i.e., not reserved)
- instance is running Elasticsearch version 7.9 and above
- instance is running on one of the following family instance types:
- M instances (m5, m4, m3)
- C instances (c5, c4)
- R instances (r5, r4, r3)
Potential savings (range in % on annual basis):
Savings could be as high as 10% on annualized OpenSearch costs based on the savings from moving to Graviton2.
What happens when the Fixer is executed?
First, the Fixer confirms the target cluster is healthy, checking the cluster state and CloudWatch health metrics from the last 24 hours. The Fixer then retypes the cluster to its Graviton2-equivalent instance type and checks to confirm the retyping process worked as expected. Once complete, all instances in the OpenSearch domain will be running on a C6g, M6g, or R6g instance, depending on the original Intel-based instance type.
Is it possible to rollback once CloudFix implements the fixer?
Yes, however, this is a manual process that requires the user to take specific actions to correct based on their configuration. There are two scenarios (both unlikely) that require rollback:
- OpenSearch domain is stuck in the “Processing” state - review this guide to manage ongoing deployments or index shard settings, and if still stuck in this state after 24 hours raise a ticket with AWS to resolve.
- OpenSearch domain has a cluster status of “Red” - identify and delete the indices marked as red, restoring them from the automated snapshots as needed. Review this guide for additional insights into cluster configuration.
Can CloudFix implement the fix automatically once I accept the recommendation?
Yes
Does this fix require downtime?
Yes, although the change is implemented using OpenSearch blue/green deployments to minimize downtime, this Fixer should run in a maintenance window.
Comments
0 comments
Article is closed for comments.