HwameiStor is a high-availability local storage system designed for cloud-native stateful workloads running on Kubernetes. Written in Go, it functions as a Cloud Native Computing Foundation sandbox project and provides persistent storage capabilities through a CSI architecture. The system creates a centralized local storage resource pool that manages various disk types including HDD, SSD, and NVMe across cluster nodes, enabling distributed services to access local volumes with data persistence guarantees.
The project comprises multiple specialized modules that work together to deliver comprehensive storage functionality. The local-disk-manager module handles disk management on individual nodes, while the local-storage module provisions high-performance persistent LVM volumes with local access. A custom scheduler automatically places pods on nodes containing their associated HwameiStor volumes, complemented by an admission controller webhook that modifies scheduler names for pods using HwameiStor volumes. The evictor module automatically migrates volumes during planned or unplanned node and pod evictions, maintaining data availability. Additional modules include an exporter for Prometheus metrics collection, volume snapshot and restore capabilities for LVM volumes, automatic volume resizing based on user-defined policies, IO throttling to prevent pod crashes under resource constraints, application failover functionality, system audit tracking, and a user interface for cluster management.
HwameiStor supports Kubernetes versions ranging from 1.18 through 1.30, with version-specific compatibility detailed in the repository. The project integrates DRBD as a third-party HA module, implementing network-based mirroring for highly available volumes. Recent feature additions include data encryption completed in version 0.16.0 and cache volumes for dataset acceleration on nodes completed in version 0.14.5.
According to GitGenius activity tracking across 182 issues and pull requests, the repository maintains a median response latency of 0.0 hours with a mean of 85.5 hours, indicating active community engagement. The most frequently tracked issue labels are kind/bug with 61 occurrences, Stale with 58, and kind/enhancement with 21. Primary contributors FloatXD, SSmallMonster, and peng9808 have logged 193, 185, and 89 tracked events respectively, demonstrating sustained development momentum. The repository shares overlapping contributors with llm-d, kubean-io, and vllm-project repositories, indicating cross-project collaboration within the cloud-native ecosystem.
The roadmap shows completed features including CSI provisioning for LVM and disk volumes, HA volume support, volume expansion and conversion, volume migration, volume groups, disk health checking, an HwameiStor operator, observability integration, failover capabilities, IO throttling, volume auto-expansion, snapshots, and cloning. Planned features include disk replacement, LVM volume thin provisioning, system consistency checks, and volume backup with remote restore capabilities. The project maintains comprehensive documentation at hwameistor.io and actively engages the community through weekly blogs, CNCF Slack channels, and WeChat discussion groups.
