Fault Injection

The fault injection filter can be used to test the resiliency of microservices to different forms of failures. The filter can be used to inject delays and abort requests with user-specified error codes, thereby providing the ability to stage different failure scenarios such as service failures, service overloads, high network latency, network partitions, etc. Faults injection can be limited to a specific set of requests based on the (destination) upstream cluster of a request and/or a set of pre-defined request headers.

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.

Configuration

Note

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

Controlling fault injection via HTTP headers

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:

  • Request delay configuration via the x-envoy-fault-delay-request header. The header value should be an integer that specifies the number of milliseconds to throttle the latency for.
  • Response rate limit configuration via the x-envoy-fault-throughput-response header. The header value should be an integer that specified the limit in KiB/s and must be > 0.

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.

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

name: envoy.fault
config:
  max_active_faults: 100
  delay:
    header_delay: {}
    percentage:
      numerator: 100
  response_rate_limit:
    header_limit: {}
    percentage:
      numerator: 100

Runtime

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 configured for abort.
fault.http.abort.http_status
HTTP status code that will be used as the 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.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 configured for delay.
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 percentage 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_delay_percent
  • fault.http.<downstream-cluster>.delay.fixed_duration_ms

Downstream cluster name is taken from the HTTP x-envoy-downstream-service-cluster header. If the following settings are not found in the runtime it defaults to the global runtime settings which defaults to the config settings.

Statistics

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

Name Type Description
delays_injected Counter Total requests that were delayed
aborts_injected Counter Total requests that were aborted
response_rl_injected Counter Total requests that had a response rate limit selected for injection (actually injection may not occur due to disconnect, reset, no body, etc.)
faults_overflow Counter Total number of faults that were not injected due to overflowing the max_active_faults setting
active_faults Gauge Total number of faults active at the current time
<downstream-cluster>.delays_injected Counter Total delayed requests for the given downstream cluster
<downstream-cluster>.aborts_injected Counter Total aborted requests for the given downstream cluster