Milvus logo

OpenEverest for Milvus

Run Milvus, the open-source vector database, on any Kubernetes cluster. Deploy a single-process standalone instance or a distributed cluster, autoscale query nodes, roll out upgrades without downtime, and back up collections to S3 through a single UI and API, powered by the Milvus Operator.

Milvus on OpenEverest
Standalone or Cluster

Standalone or Cluster

Pick a single-process standalone instance for development, or a distributed cluster with independently sized proxies, query, data, and streaming nodes - and review every setting before you create it.

Instance Overview

Instance Overview

See every component at a glance, copy the host, port, root credentials, or connection URL into pymilvus or any Milvus SDK, and edit sizing in place.

Scheduled Backups

Scheduled Backups

Back up collections on demand or on a schedule to any S3-compatible storage, with retention, and restore them into the same instance or a new one.

Visual Inspector

Visual Inspector

Follow every Milvus pod as a live graph - status, readiness, restarts, and node placement - and stream logs from the coordinator, proxies, or query nodes without leaving the UI.

Key capabilities

Standalone or Distributed

Start with a single-process instance, or run Milvus 2.6 as a cluster with an active-standby coordinator and separately scalable proxies, query, data, and streaming nodes.

Scaling & Autoscaling

Scale each component on its own and let proxies, data, and query nodes autoscale on CPU or memory. Scaling in below what a loaded collection's replicas need is refused.

Backups & Restore

On-demand and scheduled backups to S3-compatible storage with count- or time-based retention, restored into an existing instance or a brand-new one.

Rolling Upgrades

Curated Milvus 2.6 versions, upgraded component by component in the right order. Downgrades are refused, because Milvus migrates its metadata on upgrade.

Bundled or External Dependencies

Get etcd and object storage bundled and sized for you, with the write-ahead log kept in object storage (Woodpecker) - or point Milvus at your own etcd and S3-compatible bucket.

GPUs & Node Pools

Switch to GPU builds with one version change, and split query nodes across GPU pools - with their own node selectors, tolerations, and secondary networks.

Secure by Default

Authentication is always on: OpenEverest generates the root password and the storage credentials, keeps them in Kubernetes Secrets, and applies a new root password when you change it.

Powered by

milvus-io/milvus

Milvus is the open-source, cloud-native vector database built for scalable similarity search on billions of vectors.

zilliztech/milvus-operator

Milvus instances on OpenEverest are powered by the open-source Milvus Operator, which manages Milvus and its dependencies on Kubernetes.

zilliztech/milvus-backup

Backups and restores run the upstream milvus-backup tool, which copies collections and their index definitions to S3-compatible storage.

Ecosystem & extensions

Extend Milvus on OpenEverest with community plugins and integrations.

Inspector

An OpenEverest UI plugin that adds an Inspector tab to instance pages: the pods behind a Milvus instance as a live graph or a table, streamed container logs, and kubectl-style pod details.

openeverest/plugin-inspector →

Prometheus

Turn on monitoring and OpenEverest creates a PodMonitor that scrapes every Milvus component, ready for your Prometheus and Grafana dashboards.

prometheus-operator/prometheus-operator →

Frequently asked questions

Is Milvus free to run on OpenEverest?
Yes. OpenEverest, Milvus, and the Milvus Operator are all open-source with no licensing fees, and everything runs on your own Kubernetes cluster, in the cloud or on-premises.
What is a vector database used for?
A vector database stores embeddings - numeric representations of text, images, or audio - and finds the most similar ones fast. It powers retrieval-augmented generation (RAG), semantic and hybrid search, recommendations, and deduplication.
How does OpenEverest run Milvus under the hood?
The provider-milvus provider translates an OpenEverest Instance into a Milvus resource for the open-source Milvus Operator, which runs Milvus and its etcd and object storage. Read the walkthrough in Milvus on OpenEverest.
Standalone or cluster - which should I choose?
Standalone runs Milvus as a single process and is the smallest footprint, ideal for development and small datasets. The cluster topology splits Milvus into proxies, a coordinator, and query, data, and streaming nodes that scale independently, for production and larger collections.
Which Milvus versions are supported?
OpenEverest ships curated Milvus 2.6 versions, including GPU builds. Upgrades roll out component by component; downgrades are refused. See the provider repository for the current version matrix.
Do I need Pulsar or Kafka?
No. New clusters keep the write-ahead log in object storage with Woodpecker, so the only dependencies are etcd and S3-compatible storage. Pulsar is still available if you prefer it.
Can I use my own S3 bucket and etcd?
Yes. Instead of the bundled ones, point Milvus at an existing etcd cluster and an S3-compatible bucket, such as Amazon S3 or MinIO. A rootPath lets several instances share one bucket or etcd.
How do backups and restores work?
Backups use the upstream milvus-backup tool to copy collections and their index definitions to S3-compatible storage, on demand or on a schedule with retention. Restore into the same instance, or start a new instance from a backup - restores need Milvus 2.6.12 or newer.
Can Milvus use GPUs on OpenEverest?
Yes. Pick a GPU version to run the GPU build of Milvus, request GPUs per component, and use deployment groups to place query nodes on specific GPU node pools.

Ready to get started?

Deploy OpenEverest on any Kubernetes cluster and run your first Milvus vector database in minutes.

View on GitHub Read the Docs

Milvus and Zilliz are trademarks of their respective owners. OpenEverest is not affiliated with, endorsed by, or sponsored by these organizations. These names are used for identification purposes only.