Fault Injection

    The scope of failures is restricted to those that are observable by an application communicating over the network. CPU and disk failures on the local host cannot be emulated.

    Note

    The fault injection filter must be inserted before any other filter, including the router filter.

    The fault filter has the capability to allow fault configuration to be specified by the caller. This is useful in certain scenarios in which it is desired to allow the client to specify its own fault configuration. The currently supported header controls are:

    x-envoy-fault-abort-request

    HTTP status code to abort a request with. The header value should be an integer that specifies the HTTP status code to return in response to a request and must be in the range [200, 600). In order for the header to work, needs to be set.

    x-envoy-fault-abort-grpc-request

    gRPC status code to abort a request with. The header value should be a non-negative integer that specifies the gRPC status code to return in response to a request. Its value range is [0, UInt32.Max] instead of [0, 16] to allow testing even not well-defined gRPC status codes. When this header is set, the HTTP response status code will be set to 200. In order for the header to work, header_abort needs to be set. If both x-envoy-fault-abort-request and x-envoy-fault-abort-grpc-request headers are set then x-envoy-fault-abort-grpc-request header will be ignored and fault response http status code will be set to x-envoy-fault-abort-request header value.

    x-envoy-fault-abort-request-percentage

    The percentage of requests that should be failed with a status code that’s defined by the value of x-envoy-fault-abort-request or x-envoy-fault-abort-grpc-request HTTP headers. The header value should be an integer that specifies the numerator of the percentage of request to apply aborts to and must be greater or equal to 0 and its maximum value is capped by the value of the numerator of field. Percentage’s denominator is equal to default percentage’s denominator percentage field. In order for the header to work, needs to be set and either x-envoy-fault-abort-request or x-envoy-fault-abort-grpc-request HTTP header needs to be a part of the request.

    x-envoy-fault-delay-request

    The duration to delay a request by. The header value should be an integer that specifies the number of milliseconds to throttle the latency for. In order for the header to work, header_delay needs to be set.

    x-envoy-fault-delay-request-percentage

    The percentage of requests that should be delayed by a duration that’s defined by the value of x-envoy-fault-delay-request HTTP header. The header value should be an integer that specifies the percentage of request to apply delays to and must be greater or equal to 0 and its maximum value is capped by the value of the numerator of field. Percentage’s denominator is equal to default percentage’s denominator percentage field. In order for the header to work, needs to be set and x-envoy-fault-delay-request HTTP header needs to be a part of a request.

    x-envoy-fault-throughput-response

    The rate limit to use when a response to a caller is sent. The header value should be an integer that specifies the limit in KiB/s and must be > 0. In order for the header to work, header_limit needs to be set.

    x-envoy-fault-throughput-response-percentage

    The percentage of requests whose response rate should be limited to the value of x-envoy-fault-throughput-response HTTP header. The header value should be an integer that specifies the percentage of request to apply delays to and must be greater or equal to 0 and its maximum value is capped by the value of the numerator of field. Percentage’s denominator is equal to default percentage’s denominator percentage field. In order for the header to work, needs to be set and x-envoy-fault-delay-request HTTP header needs to be a part of a request.

    Attention

    Allowing header control is inherently dangerous if exposed to untrusted clients. In this case, it is suggested to use the max_active_faults setting to limit the maximum concurrent faults that can be active at any given time.

    Note

    The following is an example configuration that enables header control for both of the above options:

    The HTTP fault injection filter supports the following global runtime settings:

    Attention

    Some of the following runtime keys require the filter to be configured for the specific fault type and some do not. Please consult the documentation for each key for more information.

    fault.http.abort.abort_percent

    % of requests that will be aborted if the headers match. Defaults to the abort_percent specified in config. If the config does not contain an abort block, then abort_percent defaults to 0. For historic reasons, this runtime key is available regardless of whether the filter is .

    fault.http.abort.http_status

    HTTP status code that will be used as the response status code of requests that will be aborted if the headers match. Defaults to the HTTP status code specified in the config. If the config does not contain an abort block, then http_status defaults to 0. For historic reasons, this runtime key is available regardless of whether the filter is configured for abort.

    fault.http.abort.grpc_status

    gRPC status code that will be used as the response status code of requests that will be aborted if the headers match. Defaults to the gRPC status code specified in the config. If this field is missing from both the runtime and the config, gRPC status code in the response will be derived from fault.http.abort.http_status field. This runtime key is only available when the filter is .

    fault.http.delay.fixed_delay_percent

    % of requests that will be delayed if the headers match. Defaults to the delay_percent specified in the config or 0 otherwise. This runtime key is only available when the filter is configured for delay.

    fault.http.delay.fixed_duration_ms

    The delay duration in milliseconds. If not specified, the fixed_duration_ms specified in the config will be used. If this field is missing from both the runtime and the config, no delays will be injected. This runtime key is only available when the filter is .

    fault.http.max_active_faults

    The maximum number of active faults (of all types) that Envoy will will inject via the fault filter. This can be used in cases where it is desired that faults are 100% injected, but the user wants to avoid a situation in which too many unexpected concurrent faulting requests cause resource constraint issues. If not specified, the max_active_faults setting will be used.

    fault.http.rate_limit.response_percent

    % of requests which will have a response rate limit fault injected. Defaults to the value set in the field. This runtime key is only available when the filter is configured for response rate limiting.

    Note, fault filter runtime settings for the specific downstream cluster override the default ones if present. The following are downstream specific runtime keys:

    • fault.http.<downstream-cluster>.abort.abort_percent

    • fault.http.<downstream-cluster>.abort.http_status

    • fault.http.<downstream-cluster>.delay.fixed_duration_ms

    The fault filter outputs statistics in the http.<stat_prefix>.fault. namespace. The comes from the owning HTTP connection manager.