KalDB delivers Elasticsearch's power without the operational complexity, enabling sub-second search at 90% lower cost with S3-backed durability.
Real problems reported by engineering teams running Elasticsearch at scale
Teams spend extensive time adjusting JVM settings only to encounter different production behavior than staging environments.
View discussionBalancing shard counts proves impossible—too many harm performance, too few limit scalability.
View discussionILM policies fail silently, leaving stale indices consuming resources and budget.
View discussionCircuit breakers trip during peak traffic, triggering cascading cluster failures.
View discussionMaster node elections during high load cause indexing pauses and query timeouts.
View discussionNetwork partitions create multiple masters, leading to data corruption requiring manual intervention.
View discussionDebugging requires deep expertise. Many organizations need external consultants just to keep clusters running.
View discussionFrom Elasticsearch forums and community discussions
"JVM heap size issue...CircuitBreakingException: Data too large, data for [parent] would be larger than limit."
"ILM failing to delete closed indices...retry attempt [20077]. We've been stuck for weeks."
"Cluster red, unassigned shards, no response on writes...tried enlarging heap...with no success."
"ES overloaded during shard relocation...REST API very slow...cannot reach ES at all during peak."
Why Elasticsearch struggles with modern log workloads
Built for document search; analytics bolted on later. Not optimized for log-heavy workloads.
2010-era Java decision creates operational ceilings at scale. GC pauses are unavoidable.
Each addition (vectors, ML, security) increases complexity without improving core capabilities.
No separation prevents independent scaling. You pay for compute even when not querying.
Complexity compounds with releases. Upgrading is painful and risky.
SSPL license changes left many teams scrambling for truly open alternatives.
How KalDB addresses each Elasticsearch limitation
| Elasticsearch Problem | KalDB Solution |
|---|---|
| JVM heap OOM failures | S3-backed storage with stateless compute; no heap management required |
| Shard rebalancing storms | No shards to manage; data stored durably in S3 |
| Complex deployment (days) | Docker Compose deployment in minutes |
| Coupled compute/storage | Fully decoupled; scale query nodes independently |
| Massive infrastructure overhead | 90% lower infrastructure costs with S3 pricing |
| Performance degrades at scale | Sub-second queries at petabyte scale |
| Troubleshooting blind spots | Simplified architecture with fewer failure points |
| SSPL licensing concerns | Apache 2.0, truly open source |
S3 storage at $0.023/GB/month vs expensive EBS volumes. Pay only for active compute during queries.
Point your existing Logstash, Grafana, and Kibana to KalDB. Same queries, same dashboards, zero retraining.
Lucene-powered indexing with intelligent caching. Fast queries when you need them most.
Apache 2.0 license with no usage restrictions. No license key, no phone-home, no surprises.
Try KalDB open source today or talk to us about production deployment