MetalLB

MetalLB is a free, open-source network load balancer for Kubernetes clusters that run outside cloud-provider infrastructure. It suits operators of compatible, self-hosted Linux clusters who can manage network addressing and routing themselves.

Overview

MetalLB allocates service IP addresses from pools that the cluster operator defines, then announces those addresses on the local network. Deployment by itself does not make services reachable: MetalLB remains idle until resources such as an IPAddressPool are configured in its deployment namespace. That requirement gives experienced Kubernetes administrators precise control, but it makes address planning and configuration part of the operational workload.

The project supports two announcement models. Layer 2 mode has one cluster machine advertise a service IP, using ARP for IPv4 or NDP for IPv6. BGP mode establishes peering between cluster machines and nearby routers and can spread traffic across multiple nodes. Layer 2 is the more straightforward choice for a network where a single announcing node is acceptable; BGP is more compelling when routing control and distribution across nodes are priorities.

Key features

• Address-pool allocation: Service IPs are assigned and released from operator-managed pools. This is useful when you own the address space and want predictable allocation, but it requires maintaining valid pools as services change.
• Layer 2 announcements: ARP and NDP cover IPv4 and IPv6 networks without requiring router peering. Because one machine announces a service address, this mode favors simpler topologies over multi-node traffic distribution.
• BGP routing: Peering with nearby routers lets traffic reach multiple cluster nodes. It is the stronger option for networks already designed around BGP, though it brings routing configuration and coordination outside Kubernetes.
• FRR-K8s backend: The recommended BGP backend supports BFD, IPv6, and merging additional FRR configuration. Those capabilities matter to operators who need faster failure detection, IPv6 routing, or integration with existing FRR settings.
• Several installation paths: Kubernetes manifests, Kustomize, Helm, and an OperatorHub operator are supported. Teams can use the deployment workflow that best matches their existing cluster tooling.
• Layer 4 routing: MetalLB addresses network-level service exposure rather than acting as a broad application platform. Its value is concentrated on making Kubernetes services reachable through allocated external addresses.

Pricing

MetalLB: free (open-source project for Kubernetes load balancing). There is no paid plan, subscription, or free-trial period, and no managed service is offered. The free model is a good fit for teams prepared to operate the controller and its networking themselves; it does not include a hosted control plane or outsourced support.

Platforms

MetalLB runs on Linux and is self-hosted. Its requirements include Kubernetes 1.13.0 or later, a compatible cluster network, and IPv4 addresses. The speaker pod needs elevated permissions for its network functions, so deployment must fit your cluster's security policy. It is generally incompatible with cloud providers; AWS, Azure, DigitalOcean, and Google Cloud are listed as unsupported environments.

Who it's for

Choose MetalLB when you operate Kubernetes on your own Linux infrastructure, control the surrounding network, and need external service addresses without depending on a cloud load-balancer integration. Layer 2 suits smaller or simpler networks, while BGP is aimed at teams with router-peering expertise and a need to distribute traffic across nodes. It is not a sensible default for a cluster hosted on an unsupported public cloud, for teams wanting a managed load-balancing service, or for organizations that cannot accept configuration changes as the project evolves.

Pros and cons

Pros

• No licensing cost: The project is free and open source, so infrastructure operators can add Kubernetes load balancing without a software subscription.
• Choice of network models: ARP/NDP Layer 2 announcements cover direct local-network setups, while BGP supports router-integrated distribution across nodes.
• Flexible deployment: Manifests, Kustomize, Helm, and OperatorHub support common Kubernetes installation practices.
• Capable BGP backend: FRR-K8s adds BFD, IPv6, and merged FRR configuration for more demanding routed networks.

Cons

• Configuration is mandatory: Installation remains inactive until address pools and related resources are supplied and maintained.
• Cloud limitations: Unsupported status for AWS, Azure, DigitalOcean, and Google Cloud rules it out for many hosted clusters.
• Beta maturity: The project warns that configuration formats may change incompatibly, which raises migration risk for long-lived deployments.
• Elevated privileges: The speaker pod requires additional network permissions, a concern for tightly restricted clusters.

Alternatives

HAProxy Community Edition is a free Linux, self-hosted load balancer for TCP and HTTP proxying; choose it when proxying is the primary requirement rather than Kubernetes service-IP advertisement.

Linux Virtual Server is free, GPL-licensed Linux software and is a better direction when you need a general Linux virtual-server implementation outside MetalLB's Kubernetes-focused workflow.

PureLB is another free option for Linux, macOS, and self-hosted environments.

AWS Elastic Load Balancing is a paid cloud service with a free tier, suited to AWS-hosted applications rather than unsupported self-managed MetalLB deployments.

F5 BIG-IP APM targets paid, enterprise environments across multiple platforms and is the alternative to consider when broader commercial infrastructure capabilities are needed.

VMware Workstation Pro is free virtualization software for Linux and Windows hosts, not a direct Kubernetes load-balancer substitute.

Array APV is a paid appliance-oriented option for API, Linux, self-hosted, and web deployments.

Vultr Block Storage is paid storage rather than a load-balancing product.

Verdict

MetalLB is a strong choice for technically capable teams running self-hosted Linux Kubernetes clusters that need operator-controlled service IPs and can integrate Layer 2 or BGP announcements into their network. Its free, focused implementation is the reason to choose it. Look elsewhere if you need supported public-cloud integration, a managed service, or a more settled configuration contract than a beta project provides.