Troubleshooting load balancers
Symptoms you can hit with a load balancer, their causes and how to fix them.
The address shows nothing
Cause: The load balancer is still provisioning. If its status is error, creation failed:
the reason is shown on its page under Provisioning failed., for example a type sold out at the
provider.
Fix: Wait for active. After an error, delete it and create it again, with another type if
the type was sold out.
Nothing is forwarded
Cause: The load balancer has no service, or no target. On application, its page also warns
when no target is in the default group.
Fix: Add a service and at least one target.
A target is unhealthy but the instance is up
Cause: The health check runs against the service's destination port (or a dedicated
service's healthCheckPort) and, for http checks, needs a 2xx or 3xx on the path.
Fix: Check that something listens on that port, that the path answers 2xx or 3xx, and that the instance's firewall lets the load balancer reach it. See Health checks.
A target on lb11–lb31 shows healthy but gets no traffic
Cause: On the network types the target's status is not the health check's result.
Fix: Check the instance and the service's health check yourself.
503 from an application load balancer
Cause: The request matched a rule whose target group has no target, or no healthy target remains, or the load balancer's connection limit is reached.
Fix: Add targets to the group, fix the targets' health, or remove the rule. The Live tab counts refused connections.
421 Misdirected Request from an application load balancer
Cause: The request is for a name this load balancer does not serve, or for another load balancer's name on a connection opened for yours.
Fix: A service cannot be changed: remove the https service and add it again with the name among its domains; see Point a domain at a load balancer.
429 from an application load balancer
Cause: The load balancer is over its request rate (20,000 a second, 30,000 in the first second).
Fix: Spread the load over several load balancers. The Live tab counts throttled requests.
Adding an https service says a name is not in your hosted zones
Cause: Deutsche Cloud proves the name over DNS through its own nameservers, so the zone must be hosted under DNS.
Fix: Add the zone, or upload your own certificate; see Use your own certificate.
The certificate stays Failed
Cause: The reason is in the service's panel and in certificateError. A common one is a zone
that is hosted but not delegated to the Deutsche Cloud nameservers.
Fix: Fix it, then choose Try again; see Serve HTTPS.
https:// works, http:// is refused
Cause: The https service was added without the redirect, and no http service listens on
port 80. On an application load balancer the request is answered 421 Misdirected Request; on
the other types the connection is refused.
Fix: Remove it and add it again with the redirect, or add an http service on port 80.
The redirect cannot be turned on
Cause: Another service listens on port 80.
Fix: Remove that service first; the redirect takes port 80.
A change is refused with 409
Cause: Another service owns the load balancer: a database, cache, broker or MySQL cluster endpoint, or a Kubernetes Service. Its page says so.
Fix: Change it through its owner.
The name stopped reaching the load balancer
Cause: The load balancer was deleted, and its addresses with it.
Fix: Point the records at the new load balancer's addresses.