> For clean Markdown of any page, append .md to the page URL. For a complete documentation index, see https://ibee.ai/docs/llms.txt (full text of every page: https://ibee.ai/docs/llms-full.txt). For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://ibee.ai/docs/_mcp/server. # Load Balancers > Put one public address in front of several servers. Learn what an IBEE load balancer does, which layer and protocol to pick, which ports it listens on, what is billed, and what to check when traffic doesn't arrive. A **load balancer** is a single public address that spreads incoming traffic across several of your servers. Clients connect to the load balancer, and it forwards each request or connection to one of your servers (its **backends**). If a backend stops answering its health check, the load balancer stops sending it traffic until it recovers. Think of it as a receptionist: visitors always come to the same front desk, and the desk sends each one to a colleague who is free. ## When to use a load balancer | You want to… | Use | Why | | ------------------------------------------------------------------ | ----------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------- | | Run the same app on two or more VMs and keep it up when one fails | **Load balancer** (this page) | One address, traffic spread across healthy backends | | Give one VM a public address that you can move to another VM later | [Reserved IP](/docs/network-security/vpc-ip-management/reserved-ips) | One server, no health checks, no traffic splitting | | Open one port on a private VM in a VPC to the internet | [NAT port forwarding](/docs/network-security/vpc-ip-management#nat-port-forwarding) | NAT (network address translation) maps a port on the VPC's shared public address to the VM; no load balancer needed | | Serve images, CSS or downloads from storage, fast, worldwide | [CDN](/docs/network-security/cdn) | Caches files close to users; not for app servers | If you have only one server, you don't need a load balancer yet. Point your domain at the server's [Reserved IP](/docs/network-security/vpc-ip-management/reserved-ips) instead. ## How it works 1. **You create the load balancer** and list your backends (an IP address or hostname plus a port for each). IBEE gives it a public hostname, such as `3f9c2b7e-….lb.ibee.ai`, and a listener port. 2. **Clients connect to that hostname and port.** You can point your own domain at it with a DNS record. 3. **The load balancer checks each backend** on a schedule (a **health check**) and picks a healthy one for each request or connection, using the [routing algorithm](/docs/network-security/load-balancer/algorithms) and backend **weights**. 4. **The backend answers through the load balancer.** Your backend sees the connection coming from the load balancer, not from the client. ## Layers and protocols A load balancer works at one of two **layers**. **Layer 7 (L7)** understands web requests (HTTP), so it can route by URL path or header. **Layer 4 (L4)** only sees network connections (TCP) and forwards them without reading them. You choose the layer and protocol when you create the load balancer, and you can't change them later. | Layer | Protocol | Clients connect to | Encryption (TLS) | Use for | | ----- | ---------------------------- | --------------------------------------- | ----------------------------------------------------------- | ------------------------------------------------ | | L7 | **HTTPS** *(portal default)* | Port `443` | Ends at the load balancer, with a certificate IBEE manages | Websites and APIs | | L7 | **HTTP** | Port `80` | None | Testing, or traffic that is encrypted elsewhere | | L4 | **TCP** *(L4 default)* | A port IBEE assigns from `9000`–`19999` | None added | Databases, queues, game servers, any TCP service | | L4 | **TLS passthrough** | Port `8443` | Passed through unopened; your backend holds the certificate | End-to-end encryption | **TLS** (Transport Layer Security) is the encryption behind `https://`. See [SSL and TLS termination](/docs/network-security/load-balancer/ssl-and-tls-termination) for how each mode handles certificates. You don't pick a location for a load balancer. It runs on IBEE's shared load-balancing service and reaches your backends at the addresses you give it. ## What a load balancer can do | Feature | Where you set it | Details | | ------------------------------------------ | ----------------------- | ------------------------------------------------------------------------------------------------------------------------- | | Backends with weights | Portal and API | `1` or more backends, each with a port and a weight from `1` to `1000` | | Routing algorithm | Portal and API | Round robin, least request, random or consistent hash. See [Algorithms](/docs/network-security/load-balancer/algorithms). | | Active health checks | Portal and API | HTTP, HTTPS or TCP probes. See [Health checks](/docs/network-security/load-balancer/health-checks). | | Passive health checks | API only | Takes a backend out after repeated `5xx` errors | | Request timeout and retries | Portal and API | For HTTP and HTTPS load balancers | | Path and header rules | Portal and API, L7 only | Send `/api` to one set of backends and everything else to another | | Your own domain with automatic certificate | API only, L7 HTTPS only | See [Use your own domain](/docs/network-security/load-balancer/ssl-and-tls-termination#use-your-own-domain) | | Change settings after creation | API only | The portal can't edit a load balancer; delete and recreate it, or use `PATCH` | > **Warning** > > Three options on the create form are saved with the load balancer but **not applied to traffic yet**: **Sticky sessions**, **Proxy protocol**, and **Encrypt traffic from the load balancer to this backend**. Don't rely on them. Traffic from the load balancer to your backends is unencrypted, and your backends don't receive the client's address through proxy protocol. ## Before you begin * **A verified organization.** Until identity verification is complete, the portal shows **Verify Identity** instead of **Create Load Balancer**. See [Verify your identity](/docs/getting-started/account-setup/verify-your-identity). * **At least one backend that the load balancer can reach.** For a VM, that is its public IPv4 address. The load balancer is not inside your VPCs, so it can't reach VPC-only private addresses. See [Attach VMs](/docs/network-security/load-balancer/attach-vms). * **A firewall rule that lets traffic reach the backend port**, for both requests and health checks. See [Create firewall rules](/docs/network-security/firewalls/create-firewall-rules). * **For the API:** an [API token](/docs/tools/api-tokens) with **Load Balancer** permission (`load-balancer.read` to view, `load-balancer.write` to create, change or delete) and your workspace ID. To create one now, follow [Create a load balancer](/docs/network-security/load-balancer/create-a-load-balancer). ## Limits | Item | Limit | | --------------------------------- | ----------------------------------------------------------------------------------------- | | Load balancers per organization | Set by your plan. See [Limits and quotas](/docs/platform-fundamentals/limits-and-quotas). | | Name | 1–128 characters, unique in the workspace | | Backends | At least `1`; port `1`–`65535`; weight `1`–`1000` (default `100`) | | Request timeout | 100–300,000 ms (default 30,000 ms, which is 30 seconds) | | Health-check interval and timeout | 100–120,000 ms each (defaults 10,000 ms and 2,000 ms) | | Health-check thresholds | `1`–`20` (defaults: `2` passes to recover, `3` failures to remove) | | Retries | `1`–`10` attempts (default `3`), 100–120,000 ms per attempt (default 5,000 ms) | | Custom domains | `1` per L7 HTTPS load balancer, API only | | IP versions | Clients connect over IPv4. Backend addresses can be IPv4 or IPv6. | ## Billing Billing starts when the load balancer becomes `active` and stops when you delete it. IBEE's price list for load balancers has these items: * each hour the load balancer is active; * HTTP requests (L7); * data processed, in and out; * new connections and connection time (L4). The charges appear under **Network & Security → Load Balancer** in your billing summary. Backends are billed separately as the VMs they are; deleting a load balancer doesn't delete them. ## Troubleshooting | You see | What it means | What to do | | ----------------------------------------------------------------------- | ---------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------- | | **Verify Identity** instead of **Create Load Balancer** | Your organization isn't verified yet | Complete [identity verification](/docs/getting-started/account-setup/verify-your-identity). | | "You have reached your plan's limit for this resource." | Your plan's load balancer limit is used up | Delete a load balancer you don't need, or click **Request a limit increase**. | | An L7 endpoint returns `503`, or an L4 connection closes straight away | No backend is healthy, so there is nowhere to send traffic | See [Health checks → Troubleshooting](/docs/network-security/load-balancer/health-checks#troubleshooting). | | Your app logs show the load balancer's address instead of the visitor's | The load balancer opens its own connection to each backend | Expected. Proxy protocol, which would pass the client address, isn't applied yet. | | You need to change backends, rules or health checks | The portal has no edit option | Use the API's `PATCH` operation, or delete the load balancer and create a new one. | ## Related pages * [Create a load balancer](/docs/network-security/load-balancer/create-a-load-balancer) * [Attach VMs as backends](/docs/network-security/load-balancer/attach-vms) * [Health checks](/docs/network-security/load-balancer/health-checks) * [Algorithms](/docs/network-security/load-balancer/algorithms) * [SSL and TLS termination](/docs/network-security/load-balancer/ssl-and-tls-termination) ## Docs - [Create a load balancer](https://ibee.ai/docs/network-security/load-balancer/create-a-load-balancer.md): Create an L7 (HTTP/HTTPS) or L4 (TCP/TLS passthrough) load balancer in the portal or with the API, check that it works, change it with the API, and delete it. Includes every form default and the exact error messages. - [Algorithms](https://ibee.ai/docs/network-security/load-balancer/algorithms.md): How an IBEE load balancer picks a backend for each request — round robin, least request, random or consistent hash — how weights change the split, and how to change the algorithm with the API. - [Health checks](https://ibee.ai/docs/network-security/load-balancer/health-checks.md): How an IBEE load balancer decides a backend is healthy — the default interval (10 seconds), timeout (2 seconds) and thresholds, how long a failing backend takes to drop out and come back, and how to fix failing checks. - [SSL and TLS termination](https://ibee.ai/docs/network-security/load-balancer/ssl-and-tls-termination.md): Serve HTTPS from an IBEE load balancer with a certificate IBEE manages, put it on your own domain, or pass encrypted traffic straight to your servers. Covers ports, certificates, what is encrypted where, and the exact DNS errors. - [Attach VMs](https://ibee.ai/docs/network-security/load-balancer/attach-vms.md): Put your Cloud VMs or GPU VMs behind an IBEE load balancer — which address to use, which firewall port to open, how to add or remove a VM later with the API, and how to check that traffic reaches every VM.