Lessons Learned: Upgrading a 100 node ES Cluster from 1.7 to 7.x

Agenda:

- Challenges faced with 1.7: Mapping explosion, large cluster state, high CPU usage, EOL

- Estimating the optimal cluster size keeping performance and cost in the mind

- Cluster provisioning using ECK on AWS EKS with separate hot and cold data tiers to save costs

- Designing settings and optimal sizes of indices, aliases, and shards

- Implementing rollover and Index Lifecycle Management

- Upgrading Java/Spring based microservices using spring-data-elasticsearch and retiring TransportClient

- Data Migration from on-prem ES to ECK based ES on AWS: Direct data migration instead of snapshot-restore approach

- Performance improvement measurements