How do I configure timeouts?¶
Envoy supports a wide range of timeouts that may need to be configured depending on the deployment. This page summarizes the most important timeouts used in various scenarios.
This is not an exhaustive list of all of the configurable timeouts that Envoy supports. Depending on the deployment additional configuration may be required.
Connection timeouts apply to the entire HTTP connection and all streams the connection carries.
The HTTP protocol idle_timeout is defined in a generic message used by both the HTTP connection manager as well as upstream cluster HTTP connections. The idle timeout is the time at which a downstream or upstream connection will be terminated if there are no active streams. The default idle timeout if not otherwise specified is 1 hour. To modify the idle timeout for downstream connections use the common_http_protocol_options field in the HTTP connection manager configuration. To modify the idle timeout for upstream connections use the common_http_protocol_options field in the Cluster’s extension_protocol_options, keyed by envoy.extensions.upstreams.http.v3.HttpProtocolOptions
See below for other connection timeouts.
Stream timeouts apply to individual streams carried by an HTTP connection. Note that a stream is an HTTP/2 and HTTP/3 concept, however internally Envoy maps HTTP/1 requests to streams so in this context request/stream is interchangeable.
The HTTP connection manager request_timeout is the amount of time the connection manager will allow for the entire request stream to be received from the client.
This timeout is not enforced by default as it is not compatible with streaming requests (requests that never end). See the stream idle timeout that follows. However, if using the buffer filter, it is recommended to configure this timeout.
The HTTP connection manager request_headers_timeout determines the amount of time the client has to send only the headers on the request stream before the stream is cancelled. This can be used to prevent clients from consuming too much memory by creating large numbers of mostly-idle streams waiting for headers. The request header timeout is disabled by default.
The HTTP connection manager stream_idle_timeout is the amount of time that the connection manager will allow a stream to exist with no upstream or downstream activity. The default stream idle timeout is 5 minutes. This timeout is strongly recommended for all requests (not just streaming requests/responses) as it additionally defends against a peer that does not open the stream window once an entire response has been buffered to be sent to a downstream client.
The HTTP protocol max_stream_duration is defined in a generic message used by the HTTP connection manager. The max stream duration is the maximum time that a stream’s lifetime will span. You can use this functionality when you want to reset HTTP request/response streams periodically. You can’t use request_timeout in this situation because this timer will be disarmed if a response header is received on the request/response streams. This timeout is available on both upstream and downstream connections.
Envoy supports additional stream timeouts at the route level, as well as overriding some of the stream timeouts already introduced above.
A route timeout is the amount of time that Envoy will wait for the upstream to respond with a complete response. This timeout does not start until the entire downstream request stream has been received.
This timeout defaults to 15 seconds, however, it is not compatible with streaming responses (responses that never end), and will need to be disabled. Stream idle timeouts should be used in the case of streaming APIs as described elsewhere on this page.
The route per_try_timeout can be configured when using retries so that individual tries using a shorter timeout than the overall request timeout described above. This timeout only applies before any part of the response is sent to the downstream, which normally happens after the upstream has sent response headers. This timeout can be used with streaming endpoints to retry if the upstream fails to begin a response within the timeout.
The route MaxStreamDuration proto can be used to override the HttpConnectionManager’s max_stream_duration for individual routes as well as setting both limits and a fixed time offset on grpc-timeout headers.
In situations where envoy is under high load, Envoy can dynamically configure timeouts using scaled timeouts. Envoy supports scaled timeouts through the Overload Manager, configured in envoy bootstrap configuration. Using a reduce timeouts overload action, the Overload Manager can be configured to monitor resources and scale timeouts accordingly. For example, a common use case may be to monitor the Envoy heap size and set the scaled TimerType to HTTP_DOWNSTREAM_CONNECTION_IDLE. The overload manager will scale down the idle timeout once the scaling_threshold has been met and will set the timeout to the min timeout once the scaling_threshold is met. The full list of supported timers that can be scaled is available in the overload manager docs.
The cluster connect_timeout specifies the amount of time Envoy will wait for an upstream TCP connection to be established. If this value is not set, a default value of 5 seconds will be used.
For upstream TLS connections, the connect timeout includes the TLS handshake. For downstream connections, see below for configuration options.
The TCP proxy idle_timeout is the amount of time that the TCP proxy will allow a connection to exist with no upstream or downstream activity. The default idle timeout if not otherwise specified is 1 hour.
The transport_socket_connect_timeout specifies the amount of time Envoy will wait for a downstream client to complete transport-level negotiations. When configured on a filter chain with a TLS or ALTS transport socket, this limits the amount of time allowed to finish the encrypted handshake after establishing a TCP connection.