Longhorn vs Obstor

Longhorn is a free, open-source distributed block storage system for Kubernetes. It replicates storage across nodes and data centers to increase availability, with V1 replicated, V2 replicated and V2 sharded with erasure coding volume types. Teams can schedule recurring snapshots and backups to NFS or S3-compatible secondary storage, and create disaster recovery volumes for recovery in a second Kubernetes cluster. A GUI dashboard is included, and the management UI is described as free. The software stack can be upgraded without disrupting running storage volumes. Installation options include kubectl, Helm, Fleet, Flux and ArgoCD. Longhorn supports mTLS for securing and encrypting gRPC communication between its control and data planes, but mTLS is disabled by default. Components need root access with privileged permissions for volume operations and management. Backups require an NFSv4 client, while ReadWriteMany volumes require NFSv4.1. Longhorn runs on Linux and supports API, web and self-hosted use. It is a CNCF incubating project; SUSE Storage is described as a Longhorn-based distribution with additional support, security, lifecycle management and certified integrations.

Obstor is an Apache 2.0-licensed, S3-compatible object storage server with built-in SFTP support. It works with S3 SDKs, command-line tools, and applications without code changes. In distributed deployments, objects are sharded and replicated across connected nodes, with each object kept in at least two global locations. Its native SFTP server stores uploaded files as S3 objects. Data features include SSE-C and SSE-S3 server-side encryption with KMS integration, configurable erasure coding for drive and node failures, and object versioning with lifecycle policies. IAM controls use users, groups, policies, and service accounts with S3 and AWS IAM policy syntax. Event notifications can be sent to Kafka, NATS, AMQP, Redis, PostgreSQL, Elasticsearch, and webhooks. Installation options include a single binary, Docker, a Kubernetes Helm chart, or building from source. Standalone instructions cover macOS, GNU/Linux, and 64-bit Windows. Standalone servers are intended for development and evaluation; versioning, object locking, and bucket replication require distributed deployment with erasure coding.