gRPC retries, assuming gRPC status code is in the returned headers

docs/configuration/http_conn_man/route_config/route.rst

@@ -216,7 +216,8 @@ HTTP retry :ref:`architecture overview <arch_overview_http_routing_retry>`.
 
 retry_on
   *(required, string)* specifies the conditions under which retry takes place. These are the same
-  conditions documented for :ref:`config_http_filters_router_x-envoy-retry-on`.
+  conditions documented for :ref:`config_http_filters_router_x-envoy-retry-on` and
+  :ref:`config_http_filters_router_x-envoy-grpc-retry-on`.
 
 num_retries
   *(optional, integer)* specifies the allowed number of retries. This parameter is optional and

docs/configuration/http_filters/router_filter.rst

@@ -51,8 +51,9 @@ x-envoy-max-retries
 If a :ref:`retry policy <config_http_conn_man_route_table_route_retry>` is in place, Envoy will default to retrying one 
 time unless explicitly specified. The number of retries can be explicitly set in the 
 :ref:`route retry config <config_http_conn_man_route_table_route_retry>`  or by using this header. 
-If a :ref:`retry policy <config_http_conn_man_route_table_route_retry>` is not configured or a 
-:ref:`config_http_filters_router_x-envoy-retry-on` header is not specified, Envoy will not retry a failed request.
+If a :ref:`retry policy <config_http_conn_man_route_table_route_retry>` is not configured and
+:ref:`config_http_filters_router_x-envoy-retry-on` or
+:ref:`config_http_filters_router_x-envoy-grpc-retry-on` headers are not specified, Envoy will not retry a failed request.
 
 A few notes on how Envoy does retries:
 
@@ -120,6 +121,32 @@ Note that retry policies can also be applied at the :ref:`route level
 
 By default, Envoy will *not* perform retries unless you've configured them per above.
 
+.. _config_http_filters_router_x-envoy-grpc-retry-on:
+
+x-envoy-grpc-retry-on
+^^^^^^^^^^^^^^^^^^^^^
+Setting this header on egress requests will cause Envoy to attempt to retry failed requests (number of
+retries defaults to 1, and is configured as x-envoy-grpc-retry is, above).  gRPC retries are currently
+only supported for gRPC status codes in response headers.  gRPC status codes in trailers will not trigger
+retry logic. One or more policies can be specified  using a ',' delimited list. The supported policies are:
+
+cancelled
+  Envoy will attempt a retry if the gRPC status code in the response headers is "cancelled" (1)
+
+deadline-exceeded
+  Envoy will attempt a retry if the gRPC status code in the response headers is "deadline-exceeded" (4)
+
+resource-exhausted
+  Envoy will attempt a retry if the gRPC status code in the response headers is "resource-exhausted" (8)
+
+As with x-envoy-grpc-retry, the number of retries can be controlled via the
+:ref:`config_http_filters_router_x-envoy-max-retries` header
+
+Note that retry policies can also be applied at the :ref:`route level
+<config_http_conn_man_route_table_route_retry>`.
+
+By default, Envoy will *not* perform retries unless you've configured them per above.
+
 x-envoy-upstream-alt-stat-name
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 

include/envoy/grpc/BUILD

@@ -17,3 +17,8 @@ envoy_cc_library(
         "//include/envoy/http:header_map_interface",
     ],
 )
+
+envoy_cc_library(
+    name = "status_interface",
+    hdrs = ["status.h"],
+)

include/envoy/grpc/status.h

@@ -0,0 +1,54 @@
+#pragma once
+
+namespace Envoy {
+namespace Grpc {
+
+class Status {
+public:
+  enum GrpcStatus {
+    // The RPC completed successfully.
+    Ok = 0,
+    // The RPC was canceled.
+    Canceled = 1,
+    // Some unknown error occurred.
+    Unknown = 2,
+    // An argument to the RPC was invalid.
+    InvalidArgument = 3,
+    // The deadline for the RPC expired before the RPC completed.
+    DeadlineExceeded = 4,
+    // Some resource for the RPC was not found.
+    NotFound = 5,
+    // A resource the RPC attempted to create already exists.
+    AlreadyExists = 6,
+    // Permission was denied for the RPC.
+    PermissionDenied = 7,
+    // The RPC does not have required credentials for the RPC to succeed.
+    Unauthenticated = 16,
+    // Some resource is exhausted, resulting in RPC failure.
+    ResourceExhausted = 8,
+    // Some precondition for the RPC failed.
+    FailedPrecondition = 9,
+    // The RPC was aborted.
+    Aborted = 10,
+    // Some operation was requested outside of a legal range.
+    OutOfRange = 11,
+    // The RPC requested was not implemented.
+    Unimplemented = 12,
+    // Some internal error occurred.
+    Internal = 13,
+    // The RPC endpoint is current unavailable.
+    Unavailable = 14,
+    // There was some data loss resulting in RPC failure.
+    DataLoss = 15,
+
+    // This is a non-GRPC error code, indicating the status code in gRPC headers
+    // was missing or empty.
+    MissingStatus = -1,
+    // This is a non-GRPC error code, indicating the status code in gRPC headers
+    // was invalid.
+    InvalidCode = -2,
+  };
+};
+
+} // Grpc
+} // Envoy

include/envoy/http/header_map.h

@@ -205,6 +205,7 @@ class HeaderEntry {
   HEADER_FUNC(EnvoyMaxRetries)                                                                     \
   HEADER_FUNC(EnvoyOriginalPath)                                                                   \
   HEADER_FUNC(EnvoyRetryOn)                                                                        \
+  HEADER_FUNC(EnvoyRetryGrpcOn)                                                                    \
   HEADER_FUNC(EnvoyUpstreamAltStatName)                                                            \
   HEADER_FUNC(EnvoyUpstreamCanary)                                                                 \
   HEADER_FUNC(EnvoyUpstreamHealthCheckedCluster)                                                   \

include/envoy/router/router.h

@@ -37,10 +37,13 @@ class RedirectEntry {
 class RetryPolicy {
 public:
   // clang-format off
-  static const uint32_t RETRY_ON_5XX             = 0x1;
-  static const uint32_t RETRY_ON_CONNECT_FAILURE = 0x2;
-  static const uint32_t RETRY_ON_RETRIABLE_4XX   = 0x4;
-  static const uint32_t RETRY_ON_REFUSED_STREAM  = 0x8;
+  static const uint32_t RETRY_ON_5XX                     = 0x1;
+  static const uint32_t RETRY_ON_CONNECT_FAILURE         = 0x2;
+  static const uint32_t RETRY_ON_RETRIABLE_4XX           = 0x4;
+  static const uint32_t RETRY_ON_REFUSED_STREAM          = 0x8;
+  static const uint32_t RETRY_ON_GRPC_CANCELLED          = 0x10;
+  static const uint32_t RETRY_ON_GRPC_DEADLINE_EXCEEDED  = 0x20;
+  static const uint32_t RETRY_ON_GRPC_RESOURCE_EXHAUSTED = 0x40;
   // clang-format on
 
   virtual ~RetryPolicy() {}

source/common/grpc/BUILD

@@ -25,6 +25,7 @@ envoy_cc_library(
     external_deps = ["protobuf"],
     deps = [
         "//include/envoy/common:optional",
+        "//include/envoy/grpc:status_interface",
         "//include/envoy/http:header_map_interface",
         "//include/envoy/http:message_interface",
         "//include/envoy/stats:stats_interface",

source/common/grpc/common.cc

@@ -24,6 +24,22 @@ namespace Grpc {
 
 const std::string Common::GRPC_CONTENT_TYPE{"application/grpc"};
 
+Status::GrpcStatus Common::getGrpcStatus(const Http::HeaderMap& trailers) {
+  const Http::HeaderEntry* grpc_status_header = trailers.GrpcStatus();
+
+  uint64_t grpc_status_code;
+  if (!grpc_status_header || grpc_status_header->value().empty()) {
+    return Status::GrpcStatus::MissingStatus;
+  }
+  if (!StringUtil::atoul(grpc_status_header->value().c_str(), grpc_status_code)) {
+    return Status::GrpcStatus::InvalidCode;
+  }
+  if (grpc_status_code > Status::GrpcStatus::DataLoss) {
+    return Status::GrpcStatus::InvalidCode;
+  }
+  return static_cast<Status::GrpcStatus>(grpc_status_code);
+}
+
 void Common::chargeStat(const Upstream::ClusterInfo& cluster, const std::string& grpc_service,
                         const std::string& grpc_method, bool success) {
   cluster.statsScope()
@@ -73,13 +89,12 @@ Http::MessagePtr Common::prepareHeaders(const std::string& upstream_cluster,
 
 void Common::checkForHeaderOnlyError(Http::Message& http_response) {
   // First check for grpc-status in headers. If it is here, we have an error.
-  const Http::HeaderEntry* grpc_status_header = http_response.headers().GrpcStatus();
-  if (!grpc_status_header) {
+  Status::GrpcStatus grpc_status_code = Common::getGrpcStatus(http_response.headers());
+  if (grpc_status_code == Status::GrpcStatus::MissingStatus) {
     return;
   }
 
-  uint64_t grpc_status_code;
-  if (!StringUtil::atoul(grpc_status_header->value().c_str(), grpc_status_code)) {
+  if (grpc_status_code == Status::GrpcStatus::InvalidCode) {
     throw Exception(Optional<uint64_t>(), "bad grpc-status header");
   }
 
@@ -100,10 +115,8 @@ void Common::validateResponse(Http::Message& http_response) {
     throw Exception(Optional<uint64_t>(), "no response trailers");
   }
 
-  const Http::HeaderEntry* grpc_status_header = http_response.trailers()->GrpcStatus();
-  uint64_t grpc_status_code;
-  if (!grpc_status_header ||
-      !StringUtil::atoul(grpc_status_header->value().c_str(), grpc_status_code)) {
+  Status::GrpcStatus grpc_status_code = Common::getGrpcStatus(*http_response.trailers());
+  if (grpc_status_code < 0) {
     throw Exception(Optional<uint64_t>(), "bad grpc-status trailer");
   }
 

source/common/grpc/common.h

@@ -5,6 +5,7 @@
 
 #include "envoy/common/exception.h"
 #include "envoy/common/optional.h"
+#include "envoy/grpc/status.h"
 #include "envoy/http/header_map.h"
 #include "envoy/http/message.h"
 #include "envoy/stats/stats.h"
@@ -25,6 +26,13 @@ class Exception : public EnvoyException {
 
 class Common {
 public:
+  /**
+   * Returns the GrpcStatus code from a given set of headers, if present.
+   * @headers the headers to parse.
+   * @returns the parsed status code or InvalidCode if no valid status is found.
+   */
+  static Status::GrpcStatus getGrpcStatus(const Http::HeaderMap& headers);
+
   /**
    * Charge a success/failure stat to a cluster/service/method.
    * @param cluster supplies the target cluster.

source/common/http/headers.h

@@ -30,6 +30,7 @@ class HeaderValues {
   const LowerCaseString EnvoyMaxRetries{"x-envoy-max-retries"};
   const LowerCaseString EnvoyOriginalPath{"x-envoy-original-path"};
   const LowerCaseString EnvoyRetryOn{"x-envoy-retry-on"};
+  const LowerCaseString EnvoyRetryGrpcOn{"x-envoy-retry-grpc-on"};
   const LowerCaseString EnvoyUpstreamAltStatName{"x-envoy-upstream-alt-stat-name"};
   const LowerCaseString EnvoyUpstreamCanary{"x-envoy-upstream-canary"};
   const LowerCaseString EnvoyUpstreamRequestTimeoutAltResponse{
@@ -92,6 +93,12 @@ class HeaderValues {
     const std::string Retriable4xx{"retriable-4xx"};
   } EnvoyRetryOnValues;
 
+  struct {
+    const std::string Cancelled{"cancelled"};
+    const std::string DeadlineExceeded{"deadline-exceeded"};
+    const std::string ResourceExhausted{"resource-exhausted"};
+  } EnvoyRetryOnGrpcValues;
+
   struct {
     const std::string _100Continue{"100-continue"};
   } ExpectValues;

source/common/router/BUILD

@@ -79,6 +79,7 @@ envoy_cc_library(
         "//include/envoy/upstream:upstream_interface",
         "//source/common/common:assert_lib",
         "//source/common/common:utility_lib",
+        "//source/common/grpc:common_lib",
         "//source/common/http:codes_lib",
         "//source/common/http:headers_lib",
         "//source/common/http:utility_lib",

source/common/router/config_impl.cc

@@ -40,6 +40,8 @@ RetryPolicyImpl::RetryPolicyImpl(const Json::Object& config) {
       config.getObject("retry_policy")->getInteger("per_try_timeout_ms", 0));
   num_retries_ = config.getObject("retry_policy")->getInteger("num_retries", 1);
   retry_on_ = RetryStateImpl::parseRetryOn(config.getObject("retry_policy")->getString("retry_on"));
+  retry_on_ |=
+      RetryStateImpl::parseRetryGrpcOn(config.getObject("retry_policy")->getString("retry_on"));
 }
 
 ShadowPolicyImpl::ShadowPolicyImpl(const Json::Object& config) {

source/common/router/retry_state_impl.cc

@@ -7,6 +7,7 @@
 
 #include "common/common/assert.h"
 #include "common/common/utility.h"
+#include "common/grpc/common.h"
 #include "common/http/codes.h"
 #include "common/http/headers.h"
 #include "common/http/utility.h"
@@ -19,6 +20,9 @@ namespace Router {
 const uint32_t RetryPolicy::RETRY_ON_5XX;
 const uint32_t RetryPolicy::RETRY_ON_CONNECT_FAILURE;
 const uint32_t RetryPolicy::RETRY_ON_RETRIABLE_4XX;
+const uint32_t RetryPolicy::RETRY_ON_GRPC_CANCELLED;
+const uint32_t RetryPolicy::RETRY_ON_GRPC_DEADLINE_EXCEEDED;
+const uint32_t RetryPolicy::RETRY_ON_GRPC_RESOURCE_EXHAUSTED;
 
 RetryStatePtr RetryStateImpl::create(const RetryPolicy& route_policy,
                                      Http::HeaderMap& request_headers,
@@ -29,7 +33,8 @@ RetryStatePtr RetryStateImpl::create(const RetryPolicy& route_policy,
   RetryStatePtr ret;
 
   // We short circuit here and do not both with an allocation if there is no chance we will retry.
-  if (request_headers.EnvoyRetryOn() || route_policy.retryOn()) {
+  if (request_headers.EnvoyRetryOn() || request_headers.EnvoyRetryGrpcOn() ||
+      route_policy.retryOn()) {
     ret.reset(new RetryStateImpl(route_policy, request_headers, cluster, runtime, random,
                                  dispatcher, priority));
   }
@@ -48,12 +53,15 @@ RetryStateImpl::RetryStateImpl(const RetryPolicy& route_policy, Http::HeaderMap&
 
   if (request_headers.EnvoyRetryOn()) {
     retry_on_ = parseRetryOn(request_headers.EnvoyRetryOn()->value().c_str());
-    if (retry_on_ != 0 && request_headers.EnvoyMaxRetries()) {
-      const char* max_retries = request_headers.EnvoyMaxRetries()->value().c_str();
-      uint64_t temp;
-      if (StringUtil::atoul(max_retries, temp)) {
-        retries_remaining_ = temp;
-      }
+  }
+  if (request_headers.EnvoyRetryGrpcOn()) {
+    retry_on_ |= parseRetryGrpcOn(request_headers.EnvoyRetryGrpcOn()->value().c_str());
+  }
+  if (retry_on_ != 0 && request_headers.EnvoyMaxRetries()) {
+    const char* max_retries = request_headers.EnvoyMaxRetries()->value().c_str();
+    uint64_t temp;
+    if (StringUtil::atoul(max_retries, temp)) {
+      retries_remaining_ = temp;
     }
   }
 
@@ -96,6 +104,22 @@ uint32_t RetryStateImpl::parseRetryOn(const std::string& config) {
   return ret;
 }
 
+uint32_t RetryStateImpl::parseRetryGrpcOn(const std::string& retry_grpc_on_header) {
+  uint32_t ret = 0;
+  std::vector<std::string> retry_on_list = StringUtil::split(retry_grpc_on_header, ',');
+  for (const std::string& retry_on : retry_on_list) {
+    if (retry_on == Http::Headers::get().EnvoyRetryOnGrpcValues.Cancelled) {
+      ret |= RetryPolicy::RETRY_ON_GRPC_CANCELLED;
+    } else if (retry_on == Http::Headers::get().EnvoyRetryOnGrpcValues.DeadlineExceeded) {
+      ret |= RetryPolicy::RETRY_ON_GRPC_DEADLINE_EXCEEDED;
+    } else if (retry_on == Http::Headers::get().EnvoyRetryOnGrpcValues.ResourceExhausted) {
+      ret |= RetryPolicy::RETRY_ON_GRPC_RESOURCE_EXHAUSTED;
+    }
+  }
+
+  return ret;
+}
+
 void RetryStateImpl::resetRetry() {
   if (callback_) {
     cluster_.resourceManager(priority_).retries().dec();
@@ -169,6 +193,20 @@ bool RetryStateImpl::wouldRetry(const Http::HeaderMap* response_headers,
     }
   }
 
+  if (retry_on_ &
+          (RetryPolicy::RETRY_ON_GRPC_CANCELLED | RetryPolicy::RETRY_ON_GRPC_DEADLINE_EXCEEDED |
+           RetryPolicy::RETRY_ON_GRPC_RESOURCE_EXHAUSTED) &&
+      response_headers) {
+    Grpc::Status::GrpcStatus status = Grpc::Common::getGrpcStatus(*response_headers);
+    if ((status == Grpc::Status::Canceled && (retry_on_ & RetryPolicy::RETRY_ON_GRPC_CANCELLED)) ||
+        (status == Grpc::Status::DeadlineExceeded &&
+         (retry_on_ & RetryPolicy::RETRY_ON_GRPC_DEADLINE_EXCEEDED)) ||
+        (status == Grpc::Status::ResourceExhausted &&
+         (retry_on_ & RetryPolicy::RETRY_ON_GRPC_RESOURCE_EXHAUSTED))) {
+      return true;
+    }
+  }
+
   return false;
 }
 

source/common/router/retry_state_impl.h

@@ -27,6 +27,9 @@ class RetryStateImpl : public RetryState {
 
   static uint32_t parseRetryOn(const std::string& config);
 
+  // Returns the RetryPolicy extracted from the x-envoy-retry-grpc-on header.
+  static uint32_t parseRetryGrpcOn(const std::string& retry_grpc_on_header);
+
   // Router::RetryState
   bool enabled() override { return retry_on_ != 0; }
   bool shouldRetry(const Http::HeaderMap* response_headers,

test/common/grpc/BUILD

@@ -26,6 +26,7 @@ envoy_cc_test(
         "//source/common/http:headers_lib",
         "//test/mocks/upstream:upstream_mocks",
         "//test/proto:helloworld_proto",
+        "//test/test_common:utility_lib",
     ],
 )