Case study
ECK Upgrade for a Fortune 500 Tech Firm
How a Fortune 500 semiconductor and technology company upgraded Elastic Cloud on Kubernetes from v2.5 to v2.14, moving to vendor-maintained Helm charts.
A semiconductor and enterprise-technology company running high-volume Elasticsearch workloads across multiple Kubernetes clusters.
Its infrastructure carries more than 21TB of data per region and needed modernizing for resilience, observability, and scale.
ECK upgraded from v2.5 onto vendor-maintained Helm charts. Large-memory nodes tuned and benchmarked with Elasticsearch Rally. Per-region data carried through a phased, snapshot-validated migration.
Key servicesHow SquareShift delivered it.
The challenge
The client — a semiconductor and enterprise-technology company running Elasticsearch across multiple Kubernetes clusters, with more than 21TB of data per region — was still managing its Elastic Cloud on Kubernetes deployment with custom Helm charts on the aging ECK 2.5.
Every large-memory node, every cross-region cluster, and every custom Helm value had to be re-validated by hand. The client wanted an upgrade to ECK 2.14 that moved onto vendor-maintained Helm, tightened resilience, and gave the team real observability inside Kubernetes.
What we delivered
SquareShift planned and tested the ECK upgrade path with snapshot validation at every stage, then moved the client off custom Helm charts and onto Elastic’s official, vendor-maintained charts — tuning node roles, JVM settings, and memory configuration for nodes running up to 128GB.
Performance was benchmarked with Elasticsearch Rally before and after the change, ILM hot-warm-cold tiering, CCS, CCR, and rack awareness were implemented for resilience, and Filebeat, Elastic APM, and Fleet were enabled to give the team full-stack observability inside Kubernetes.
The payoff
The client’s Elasticsearch workloads now run on ECK 2.14 with vendor-maintained Helm charts, replacing the custom configuration the team previously had to patch and validate on its own.
Large-memory nodes are tuned and Rally-benchmarked, more than 21TB of data per region moved through a snapshot-validated migration path, and Filebeat/APM/Fleet observability now runs natively inside Kubernetes.
Hand-rolled Helm charts feel like control until the vendor ships a change you patch around alone — vendor-maintained charts trade some flexibility for far less on-call risk.
Cloud Modernization Practice Lead, SquareShift
Where this work sits