tikv/pd

PD is a placement driver and cluster manager for TiKV, the distributed key-value store.

PD solves the problem of coordinating and scheduling a distributed TiKV cluster. It manages cluster topology, data placement, and scheduling decisions across nodes. The tool embeds etcd to provide fault tolerance and consensus, ensuring that cluster state remains consistent even when individual nodes fail. PD exposes an HTTP API for querying cluster membership and state, and provides command-line tools for administration and recovery.

PD is essential if you are deploying TiKV as a distributed storage system. It is a required component rather than an optional add-on; TiKV clusters cannot function without a running PD instance. The tool can be deployed as a single node for development or testing, or as a multi-node cluster for production use. PD integrates with the broader TiKV ecosystem, which includes TiDB for SQL access and TiUP for cluster deployment automation. Docker images are available for containerized deployments, and Kubernetes support is documented for cloud-native environments.

The project maintains active development with regular updates to its codebase. Contributions are actively solicited through documented contribution guidelines and a contribution map. The build process is straightforward, requiring only Go and a make invocation to produce the server binary, control utility, and recovery tool.