$ k8scalc compare loki elasticsearch
Loki vs Elasticsearch for Kubernetes Logging
Loki vs Elasticsearch (ELK/ECK) for Kubernetes log aggregation — compare storage cost, RAM requirements, query capabilities, and Grafana integration.
FeatureLokiElasticsearch
Storage approach
Label index + chunks
Full-text inverted index
Storage cost
Low (5–10× less)
High
RAM per node
~500 MB–2 GB
2–8 GB
Full-text search
No (label-based only)
Yes
Grafana integration
Native
Via plugin
LogQL query language
Yes (LogQL)—
Kibana / ES DSL—
Setup complexity
Low
High (ECK operator)
S3 / object storage backend
Yes (native)
Snapshots only
Log schema requirements
Labels only
Any structured data
CNCF project
Yes
No
Verdict
Loki is the right choice for most Kubernetes clusters. It integrates natively with Grafana, uses 5–10× less storage than Elasticsearch (no full-text index on all fields), and has much lower RAM requirements. Elasticsearch wins if you need full-text search across log content, complex aggregations, or already have a centralized ELK stack. For a self-hosted Hetzner cluster where disk cost and RAM are real constraints, Loki with S3 backend (Hetzner Object Storage) is the cost-effective default.