FAQs
For more detailed information about Load Balancing — including how-to guides, tutorials, and other reference information — check out our product documentation.
This issue may be caused by a combination of two issues.
When you attach a monitor to a pool, you can specify the Health Monitor Regions that Cloudflare uses to monitor your endpoint health.
If you select multiple regions or choose All Data Centers (Enterprise Only), you may dramatically increase traffic to that pool and its associated endpoints. Each region sends individual health monitor requests from 3 data centers. Using All Data Centers sends individual health monitor requests from all existing Cloudflare data centers (and that number of data centers is growing all the time).
To reduce traffic, reduce the number of selected regions or choose an option besides All Data Centers.
If you have a low interval for your health monitor requests, you may increase the traffic sent to your endpoints.
To learn more about how endpoints and pools become unhealthy, refer to Endpoint and pool health.
If you know that your endpoint is healthy but load balancing is reporting it as unhealthy, check the following settings on the monitor:
- Perform a
curlrequest against the configured endpoint. Make sure the response you are seeing matches your settings for the monitor. - Ensure your firewall or web server does not block or rate limit our health monitors and accepts requests from Cloudflare IP addresses.
- If you are looking for a specific value in the Response Body, make sure that value is relatively static and within the first 10 KB of the HTML page.
- If your endpoint responds with a
301or302status code, make sure Follow Redirects is selected. - Try increasing the Timeout value.
- Review the Host Header for the health monitor.
- If you are using Authenticated Origin Pulls, Argo Smart Routing, make sure that you entered a zone value for Simulate Zone.
You occasionally might see traffic routed away from a pool if a health monitor request fails from a specific data center (even if the endpoint is still healthy). That data center may direct a small number of requests to another pool that is considered healthy by that data center.
To learn more about how endpoints and pools become unhealthy, refer to Endpoint and pool health.
When a pool or endpoint becomes unhealthy, traffic may be rerouted to other healthy pools or endpoints based on your configuration. You might experience this behavior when using:
1 - Pools with All-Datacenters monitoring and the monitor fails in a specific data center. In this case, all traffic will be steered away from impacted endpoints in that datacenter until the monitor succeeds again. These instances are reflected in LB request analytics as steering away from an unhealthy endpoint or pool.
2 - Pools with FQDN endpoint addresses and the recursive DNS lookup fails in a specific data center. In this case, only requests for which the DNS request fails will be steered away from impacted endpoints. This could be sporadic, especially if upstream authoritative resolvers occasionally time out or fail, when the local DNS cache TTL expires and a remote lookup is required in the hot path. This also appears in LB request analytics as steering away from an unhealthy endpoint, and the resolved endpoint IP will be missing from the request log.
To avoid these scenarios:
1 - Do not use All-Datacenters monitoring.
2 - Use IP addresses for endpoint configurations. If that is not feasible, use domains for which Cloudflare is authoritative (primary or secondary).
To learn more about how endpoints and pools become unhealthy, refer to Endpoint and pool health.
Cloudflare Load Balancing helps monitor endpoints health and — based on that and other information — route incoming requests accordingly. Individual endpoints have monitors attached, which issue monitor requests at regular intervals.
Cloudflare Health Checks are identical to monitors within a load balancer, but only meant for probing server health (and not distributing traffic).
You may see different numbers of requests when reviewing Load Balancing Analytics, especially when compared to other Cloudflare dashboards (Caching, etc.).
Load balancing requests are the number of uncached requests made by your load balancer. By default, Cloudflare caches resolved IP addresses for up to five seconds. This built-in caching is often the cause of an discrepancies.
For a list of specific error codes and next steps, refer to Load Balancing Troubleshooting.