Elastic APM agents can send unauthenticated (anonymous) events to the APM Server. An event is considered to be anonymous if no authentication token can be extracted from the incoming request. The APM Server’s default response to these these requests depends on its configuration:
Anonymous requests are rejected and an authentication error is returned.
No API key or secret token is configured
Anonymous requests are accepted by the APM Server.
In some cases, however, it makes sense to allow both authenticated and anonymous requests. For example, it isn’t possible to authenticate requests from front-end services as the secret token or API key can’t be protected. This is the case with the Real User Monitoring (RUM) agent running in a browser, or the iOS/Swift agent running in a user application. However, you still likely want to authenticate requests from back-end services. To solve this problem, you can enable anonymous authentication in the APM Server to allow the ingestion of unauthenticated client-side APM data while still requiring authentication for server-side services.
Configuring anonymous auth for client-side servicesedit
When configuring anonymous authentication for client-side services, there are a few configuration variables that can mitigate the impact of malicious requests to an unauthenticated APM Server endpoint.
Use the Allowed anonymous agents and Allowed anonymous services configs to ensure that the
service.name of each incoming request match a specified list.
Additionally, the APM Server can rate-limit unauthenticated requests based on the client IP address
client.ip) of the request.
This allows you to specify the maximum number of requests allowed per unique IP address, per second.
When an API key or secret token is configured,
anonymous authentication must be enabled to collect RUM data.
To enable anonymous access, set either
Because anyone can send anonymous events to the APM Server,
additional configuration variables are available to rate limit the number anonymous events the APM Server processes;
throughput is equal to the
rate_limit.ip_limit times the
See Anonymous authentication for a complete list of options and a sample configuration file.
Deriving an incoming request’s
The remote IP address of an incoming request might be different from the end-user’s actual IP address, for example, because of a proxy. For this reason, the APM Server attempts to derive the IP address of an incoming request from HTTP headers. The supported headers are parsed in the following order:
If none of these headers are present, the remote address for the incoming request is used.
Using a reverse proxy or load balanceredit
HTTP headers are easily modified;
it’s possible for anyone to spoof the derived
client.ip value by changing or setting,
for example, the value of the
For this reason, if any of your clients are not trusted,
we recommend setting up a reverse proxy or load balancer in front of the APM Server.
Using a proxy allows you to clear any existing IP-forwarding HTTP headers, and replace them with one set by the proxy. This prevents malicious users from cycling spoofed IP addresses to bypass the APM Server’s rate limiting feature.