Does Langfuse's specific ClickHouse usage require RWX or RWO?

[stahamtan]

Describe your question

According to Bitnami Clickhouse docs, it needs ReadWriteMany volumes for deployment scaling. However, in my self-hosted langfuse, all PVCs including those for Clickhouse are ReadWriteOnce. I am wondering whether this is fine for langfuse's specific usage of Clickhouse?

Link to the docs: https://github.com/bitnami/charts/tree/de0ac1461d486cd15f4a0a3eed8cf75ffcd55995/bitnami/clickhouse#prerequisites

Langfuse Cloud or Self-Hosted?

Self-Hosted

If Self-Hosted

1.5.15

If Langfuse Cloud

No response

SDK and integration versions

No response

Pre-Submission Checklist

• I have checked for existing issues/discussions and consulted Langfuse AI.

[inkeep-ai-assistant[bot]]

Based on the available documentation, Langfuse's specific ClickHouse usage with ReadWriteOnce (RWO) volumes should work correctly for your deployment.

The Langfuse documentation for Kubernetes deployment recommends a minimum production setup with replicaCount: 3 for ClickHouse(1). The documentation states that "The ClickHouse Helm chart uses the default storage class to create volumes for each replica"(1). It also notes that you should "Ensure that the storage class has allowVolumeExpansion = true as observability workloads tend to be very disk heavy"(1).

Importantly, the documentation does not explicitly require ReadWriteMany (RWX) access mode for ClickHouse volumes in Langfuse deployments(1). The Bitnami ClickHouse Helm Chart is used as a dependency for Langfuse K8s deployments(1), and the standard configuration creates individual volumes for each replica using the default storage class(1).

Since ClickHouse replicas each maintain their own data and use replication to synchronize between instances, ReadWriteOnce volumes (one volume per pod) are sufficient. Each ClickHouse replica pod writes to its own dedicated volume, which aligns with the RWO access pattern.

Your current setup with RWO volumes should be fine for Langfuse's specific usage of ClickHouse, as long as each replica has its own persistent volume and replication is properly configured.

📚 Sources:

1. Kubernetes (Helm) (self-hosted) - Langfuse

---

Have another question? Just tag @inkeep.

[jannikmaierhoefer]

Hi @stahamtan, was inkeep able to help you here?