grpc 1.43.1 → 1.48.0

data/src/core/ext/filters/client_channel/lb_policy/xds/xds_cluster_impl.cc

@@ -16,11 +16,29 @@
 
 #include <grpc/support/port_platform.h>
 
-#include <atomic>
+#include <stddef.h>
+#include <stdint.h>
 
+#include <algorithm>
+#include <atomic>
+#include <map>
+#include <memory>
+#include <string>
+#include <utility>
+#include <vector>
+
+#include "absl/base/thread_annotations.h"
+#include "absl/memory/memory.h"
+#include "absl/status/status.h"
+#include "absl/status/statusor.h"
+#include "absl/strings/str_cat.h"
 #include "absl/strings/string_view.h"
+#include "absl/types/optional.h"
+#include "absl/types/variant.h"
 
-#include <grpc/grpc.h>
+#include <grpc/impl/codegen/connectivity_state.h>
+#include <grpc/impl/codegen/grpc_types.h>
+#include <grpc/support/log.h>
 
 #include "src/core/ext/filters/client_channel/lb_policy.h"
 #include "src/core/ext/filters/client_channel/lb_policy/child_policy_handler.h"
@@ -28,15 +46,24 @@
 #include "src/core/ext/filters/client_channel/lb_policy/xds/xds_channel_args.h"
 #include "src/core/ext/filters/client_channel/lb_policy_factory.h"
 #include "src/core/ext/filters/client_channel/lb_policy_registry.h"
+#include "src/core/ext/filters/client_channel/subchannel_interface.h"
+#include "src/core/ext/xds/xds_bootstrap.h"
 #include "src/core/ext/xds/xds_client.h"
 #include "src/core/ext/xds/xds_client_stats.h"
+#include "src/core/ext/xds/xds_endpoint.h"
 #include "src/core/lib/channel/channel_args.h"
-#include "src/core/lib/gpr/env.h"
+#include "src/core/lib/debug/trace.h"
 #include "src/core/lib/gpr/string.h"
+#include "src/core/lib/gprpp/debug_location.h"
 #include "src/core/lib/gprpp/orphanable.h"
+#include "src/core/lib/gprpp/ref_counted.h"
 #include "src/core/lib/gprpp/ref_counted_ptr.h"
 #include "src/core/lib/gprpp/sync.h"
-#include "src/core/lib/iomgr/work_serializer.h"
+#include "src/core/lib/iomgr/error.h"
+#include "src/core/lib/iomgr/pollset_set.h"
+#include "src/core/lib/json/json.h"
+#include "src/core/lib/resolver/server_address.h"
+#include "src/core/lib/transport/connectivity_state.h"
 
 namespace grpc_core {
 
@@ -118,14 +145,13 @@ class XdsClusterImplLbConfig : public LoadBalancingPolicy::Config {
   XdsClusterImplLbConfig(
       RefCountedPtr<LoadBalancingPolicy::Config> child_policy,
       std::string cluster_name, std::string eds_service_name,
-      absl::optional<std::string> lrs_load_reporting_server_name,
+      absl::optional<XdsBootstrap::XdsServer> lrs_load_reporting_server,
       uint32_t max_concurrent_requests,
-      RefCountedPtr<XdsApi::EdsUpdate::DropConfig> drop_config)
+      RefCountedPtr<XdsEndpointResource::DropConfig> drop_config)
       : child_policy_(std::move(child_policy)),
         cluster_name_(std::move(cluster_name)),
         eds_service_name_(std::move(eds_service_name)),
-        lrs_load_reporting_server_name_(
-            std::move(lrs_load_reporting_server_name)),
+        lrs_load_reporting_server_(std::move(lrs_load_reporting_server)),
         max_concurrent_requests_(max_concurrent_requests),
         drop_config_(std::move(drop_config)) {}
 
@@ -136,11 +162,12 @@ class XdsClusterImplLbConfig : public LoadBalancingPolicy::Config {
   }
   const std::string& cluster_name() const { return cluster_name_; }
   const std::string& eds_service_name() const { return eds_service_name_; }
-  const absl::optional<std::string>& lrs_load_reporting_server_name() const {
-    return lrs_load_reporting_server_name_;
+  const absl::optional<XdsBootstrap::XdsServer>& lrs_load_reporting_server()
+      const {
+    return lrs_load_reporting_server_;
   };
   uint32_t max_concurrent_requests() const { return max_concurrent_requests_; }
-  RefCountedPtr<XdsApi::EdsUpdate::DropConfig> drop_config() const {
+  RefCountedPtr<XdsEndpointResource::DropConfig> drop_config() const {
     return drop_config_;
   }
 
@@ -148,9 +175,9 @@ class XdsClusterImplLbConfig : public LoadBalancingPolicy::Config {
   RefCountedPtr<LoadBalancingPolicy::Config> child_policy_;
   std::string cluster_name_;
   std::string eds_service_name_;
-  absl::optional<std::string> lrs_load_reporting_server_name_;
+  absl::optional<XdsBootstrap::XdsServer> lrs_load_reporting_server_;
   uint32_t max_concurrent_requests_;
-  RefCountedPtr<XdsApi::EdsUpdate::DropConfig> drop_config_;
+  RefCountedPtr<XdsEndpointResource::DropConfig> drop_config_;
 };
 
 // xDS Cluster Impl LB policy.
@@ -205,7 +232,7 @@ class XdsClusterImplLb : public LoadBalancingPolicy {
 
     RefCountedPtr<CircuitBreakerCallCounterMap::CallCounter> call_counter_;
     uint32_t max_concurrent_requests_;
-    RefCountedPtr<XdsApi::EdsUpdate::DropConfig> drop_config_;
+    RefCountedPtr<XdsEndpointResource::DropConfig> drop_config_;
     RefCountedPtr<XdsClusterDropStats> drop_stats_;
     RefCountedPtr<RefCountedPicker> picker_;
   };
@@ -238,7 +265,7 @@ class XdsClusterImplLb : public LoadBalancingPolicy {
 
   OrphanablePtr<LoadBalancingPolicy> CreateChildPolicyLocked(
       const grpc_channel_args* args);
-  void UpdateChildPolicyLocked(ServerAddressList addresses,
+  void UpdateChildPolicyLocked(absl::StatusOr<ServerAddressList> addresses,
                                const grpc_channel_args* args);
 
   void MaybeUpdatePickerLocked();
@@ -461,21 +488,30 @@ void XdsClusterImplLb::UpdateLocked(UpdateArgs args) {
   config_ = std::move(args.config);
   // On initial update, create drop stats.
   if (is_initial_update) {
-    if (config_->lrs_load_reporting_server_name().has_value()) {
+    if (config_->lrs_load_reporting_server().has_value()) {
       drop_stats_ = xds_client_->AddClusterDropStats(
-          config_->lrs_load_reporting_server_name().value(),
-          config_->cluster_name(), config_->eds_service_name());
+          config_->lrs_load_reporting_server().value(), config_->cluster_name(),
+          config_->eds_service_name());
+      if (drop_stats_ == nullptr) {
+        gpr_log(GPR_ERROR,
+                "[xds_cluster_impl_lb %p] Failed to get cluster drop stats for "
+                "LRS server %s, cluster %s, EDS service name %s, load "
+                "reporting for drops will not be done.",
+                this, config_->lrs_load_reporting_server()->server_uri.c_str(),
+                config_->cluster_name().c_str(),
+                config_->eds_service_name().c_str());
+      }
     }
     call_counter_ = g_call_counter_map->GetOrCreate(
         config_->cluster_name(), config_->eds_service_name());
   } else {
     // Cluster name, EDS service name, and LRS server name should never
-    // change, because the EDS policy above us should be swapped out if
-    // that happens.
+    // change, because the xds_cluster_resolver policy above us should be
+    // swapped out if that happens.
     GPR_ASSERT(config_->cluster_name() == old_config->cluster_name());
     GPR_ASSERT(config_->eds_service_name() == old_config->eds_service_name());
-    GPR_ASSERT(config_->lrs_load_reporting_server_name() ==
-               old_config->lrs_load_reporting_server_name());
+    GPR_ASSERT(config_->lrs_load_reporting_server() ==
+               old_config->lrs_load_reporting_server());
   }
   // Update picker if max_concurrent_requests has changed.
   if (is_initial_update || config_->max_concurrent_requests() !=
@@ -494,8 +530,7 @@ void XdsClusterImplLb::MaybeUpdatePickerLocked() {
     if (GRPC_TRACE_FLAG_ENABLED(grpc_xds_cluster_impl_lb_trace)) {
       gpr_log(GPR_INFO,
               "[xds_cluster_impl_lb %p] updating connectivity (drop all): "
-              "state=READY "
-              "picker=%p",
+              "state=READY picker=%p",
               this, drop_picker.get());
     }
     channel_control_helper()->UpdateState(GRPC_CHANNEL_READY, absl::Status(),
@@ -508,8 +543,7 @@ void XdsClusterImplLb::MaybeUpdatePickerLocked() {
     if (GRPC_TRACE_FLAG_ENABLED(grpc_xds_cluster_impl_lb_trace)) {
       gpr_log(GPR_INFO,
               "[xds_cluster_impl_lb %p] updating connectivity: state=%s "
-              "status=(%s) "
-              "picker=%p",
+              "status=(%s) picker=%p",
               this, ConnectivityStateName(state_), status_.ToString().c_str(),
               drop_picker.get());
     }
@@ -541,8 +575,9 @@ OrphanablePtr<LoadBalancingPolicy> XdsClusterImplLb::CreateChildPolicyLocked(
   return lb_policy;
 }
 
-void XdsClusterImplLb::UpdateChildPolicyLocked(ServerAddressList addresses,
-                                               const grpc_channel_args* args) {
+void XdsClusterImplLb::UpdateChildPolicyLocked(
+    absl::StatusOr<ServerAddressList> addresses,
+    const grpc_channel_args* args) {
   // Create policy if needed.
   if (child_policy_ == nullptr) {
     child_policy_ = CreateChildPolicyLocked(args);
@@ -573,7 +608,7 @@ RefCountedPtr<SubchannelInterface> XdsClusterImplLb::Helper::CreateSubchannel(
   if (xds_cluster_impl_policy_->shutting_down_) return nullptr;
   // If load reporting is enabled, wrap the subchannel such that it
   // includes the locality stats object, which will be used by the EdsPicker.
-  if (xds_cluster_impl_policy_->config_->lrs_load_reporting_server_name()
+  if (xds_cluster_impl_policy_->config_->lrs_load_reporting_server()
           .has_value()) {
     RefCountedPtr<XdsLocalityName> locality_name;
     auto* attribute = address.GetAttribute(kXdsLocalityNameAttributeKey);
@@ -584,15 +619,26 @@ RefCountedPtr<SubchannelInterface> XdsClusterImplLb::Helper::CreateSubchannel(
     }
     RefCountedPtr<XdsClusterLocalityStats> locality_stats =
         xds_cluster_impl_policy_->xds_client_->AddClusterLocalityStats(
-            *xds_cluster_impl_policy_->config_
-                 ->lrs_load_reporting_server_name(),
+            xds_cluster_impl_policy_->config_->lrs_load_reporting_server()
+                .value(),
             xds_cluster_impl_policy_->config_->cluster_name(),
             xds_cluster_impl_policy_->config_->eds_service_name(),
             std::move(locality_name));
-    return MakeRefCounted<StatsSubchannelWrapper>(
-        xds_cluster_impl_policy_->channel_control_helper()->CreateSubchannel(
-            std::move(address), args),
-        std::move(locality_stats));
+    if (locality_stats != nullptr) {
+      return MakeRefCounted<StatsSubchannelWrapper>(
+          xds_cluster_impl_policy_->channel_control_helper()->CreateSubchannel(
+              std::move(address), args),
+          std::move(locality_stats));
+    }
+    gpr_log(GPR_ERROR,
+            "[xds_cluster_impl_lb %p] Failed to get locality stats object for "
+            "LRS server %s, cluster %s, EDS service name %s; load reports will "
+            "not be generated (not wrapping subchannel)",
+            this,
+            xds_cluster_impl_policy_->config_->lrs_load_reporting_server()
+                ->server_uri.c_str(),
+            xds_cluster_impl_policy_->config_->cluster_name().c_str(),
+            xds_cluster_impl_policy_->config_->eds_service_name().c_str());
   }
   // Load reporting not enabled, so don't wrap the subchannel.
   return xds_cluster_impl_policy_->channel_control_helper()->CreateSubchannel(
@@ -660,7 +706,7 @@ class XdsClusterImplLbFactory : public LoadBalancingPolicyFactory {
 
   RefCountedPtr<LoadBalancingPolicy::Config> ParseLoadBalancingConfig(
       const Json& json, grpc_error_handle* error) const override {
-    GPR_DEBUG_ASSERT(error != nullptr && *error == GRPC_ERROR_NONE);
+    GPR_DEBUG_ASSERT(error != nullptr && GRPC_ERROR_IS_NONE(*error));
     if (json.type() == Json::Type::JSON_NULL) {
       // This policy was configured in the deprecated loadBalancingPolicy
       // field or in the client API.
@@ -682,7 +728,7 @@ class XdsClusterImplLbFactory : public LoadBalancingPolicyFactory {
       child_policy = LoadBalancingPolicyRegistry::ParseLoadBalancingConfig(
           it->second, &parse_error);
       if (child_policy == nullptr) {
-        GPR_DEBUG_ASSERT(parse_error != GRPC_ERROR_NONE);
+        GPR_DEBUG_ASSERT(!GRPC_ERROR_IS_NONE(parse_error));
         std::vector<grpc_error_handle> child_errors;
         child_errors.push_back(parse_error);
         error_list.push_back(
@@ -713,14 +759,21 @@ class XdsClusterImplLbFactory : public LoadBalancingPolicyFactory {
       }
     }
     // LRS load reporting server name.
-    absl::optional<std::string> lrs_load_reporting_server_name;
-    it = json.object_value().find("lrsLoadReportingServerName");
+    absl::optional<XdsBootstrap::XdsServer> lrs_load_reporting_server;
+    it = json.object_value().find("lrsLoadReportingServer");
     if (it != json.object_value().end()) {
-      if (it->second.type() != Json::Type::STRING) {
+      if (it->second.type() != Json::Type::OBJECT) {
         error_list.push_back(GRPC_ERROR_CREATE_FROM_STATIC_STRING(
-            "field:lrsLoadReportingServerName error:type should be string"));
+            "field:lrsLoadReportingServer error:type should be object"));
       } else {
-        lrs_load_reporting_server_name = it->second.string_value();
+        grpc_error_handle parser_error;
+        lrs_load_reporting_server = XdsBootstrap::XdsServer::Parse(
+            it->second.object_value(), &parser_error);
+        if (!GRPC_ERROR_IS_NONE(parser_error)) {
+          error_list.push_back(GRPC_ERROR_CREATE_FROM_CPP_STRING(
+              absl::StrCat("errors parsing lrs_load_reporting_server")));
+          error_list.push_back(parser_error);
+        }
       }
     }
     // Max concurrent requests.
@@ -736,7 +789,7 @@ class XdsClusterImplLbFactory : public LoadBalancingPolicyFactory {
       }
     }
     // Drop config.
-    auto drop_config = MakeRefCounted<XdsApi::EdsUpdate::DropConfig>();
+    auto drop_config = MakeRefCounted<XdsEndpointResource::DropConfig>();
     it = json.object_value().find("dropCategories");
     if (it == json.object_value().end()) {
       error_list.push_back(GRPC_ERROR_CREATE_FROM_STATIC_STRING(
@@ -756,13 +809,13 @@ class XdsClusterImplLbFactory : public LoadBalancingPolicyFactory {
     }
     return MakeRefCounted<XdsClusterImplLbConfig>(
         std::move(child_policy), std::move(cluster_name),
-        std::move(eds_service_name), std::move(lrs_load_reporting_server_name),
+        std::move(eds_service_name), std::move(lrs_load_reporting_server),
         max_concurrent_requests, std::move(drop_config));
   }
 
  private:
   static std::vector<grpc_error_handle> ParseDropCategories(
-      const Json& json, XdsApi::EdsUpdate::DropConfig* drop_config) {
+      const Json& json, XdsEndpointResource::DropConfig* drop_config) {
     std::vector<grpc_error_handle> error_list;
     if (json.type() != Json::Type::ARRAY) {
       error_list.push_back(GRPC_ERROR_CREATE_FROM_STATIC_STRING(
@@ -786,7 +839,7 @@ class XdsClusterImplLbFactory : public LoadBalancingPolicyFactory {
   }
 
   static std::vector<grpc_error_handle> ParseDropCategory(
-      const Json& json, XdsApi::EdsUpdate::DropConfig* drop_config) {
+      const Json& json, XdsEndpointResource::DropConfig* drop_config) {
     std::vector<grpc_error_handle> error_list;
     if (json.type() != Json::Type::OBJECT) {
       error_list.push_back(GRPC_ERROR_CREATE_FROM_STATIC_STRING(

data/src/core/ext/filters/client_channel/lb_policy/xds/xds_cluster_manager.cc

@@ -16,28 +16,49 @@
 
 #include <grpc/support/port_platform.h>
 
+#include <stddef.h>
+
+#include <algorithm>
+#include <map>
+#include <memory>
 #include <set>
 #include <string>
+#include <utility>
 #include <vector>
 
+#include "absl/memory/memory.h"
 #include "absl/status/status.h"
+#include "absl/status/statusor.h"
 #include "absl/strings/str_cat.h"
 #include "absl/strings/string_view.h"
 
-#include <grpc/grpc.h>
+#include <grpc/impl/codegen/connectivity_state.h>
+#include <grpc/impl/codegen/grpc_types.h>
+#include <grpc/support/log.h>
 
+#include "src/core/ext/filters/client_channel/client_channel.h"
 #include "src/core/ext/filters/client_channel/lb_policy.h"
 #include "src/core/ext/filters/client_channel/lb_policy/child_policy_handler.h"
 #include "src/core/ext/filters/client_channel/lb_policy_factory.h"
 #include "src/core/ext/filters/client_channel/lb_policy_registry.h"
 #include "src/core/ext/filters/client_channel/resolver/xds/xds_resolver.h"
+#include "src/core/ext/filters/client_channel/subchannel_interface.h"
 #include "src/core/lib/channel/channel_args.h"
-#include "src/core/lib/gpr/string.h"
+#include "src/core/lib/debug/trace.h"
+#include "src/core/lib/gprpp/debug_location.h"
 #include "src/core/lib/gprpp/orphanable.h"
+#include "src/core/lib/gprpp/ref_counted.h"
 #include "src/core/lib/gprpp/ref_counted_ptr.h"
+#include "src/core/lib/gprpp/time.h"
+#include "src/core/lib/iomgr/closure.h"
+#include "src/core/lib/iomgr/error.h"
+#include "src/core/lib/iomgr/exec_ctx.h"
+#include "src/core/lib/iomgr/pollset_set.h"
 #include "src/core/lib/iomgr/timer.h"
 #include "src/core/lib/iomgr/work_serializer.h"
-#include "src/core/lib/transport/error_utils.h"
+#include "src/core/lib/json/json.h"
+#include "src/core/lib/resolver/server_address.h"
+#include "src/core/lib/transport/connectivity_state.h"
 
 #define GRPC_XDS_CLUSTER_MANAGER_CHILD_RETENTION_INTERVAL_MS (15 * 60 * 1000)
 
@@ -122,7 +143,7 @@ class XdsClusterManagerLb : public LoadBalancingPolicy {
     void Orphan() override;
 
     void UpdateLocked(RefCountedPtr<LoadBalancingPolicy::Config> config,
-                      const ServerAddressList& addresses,
+                      const absl::StatusOr<ServerAddressList>& addresses,
                       const grpc_channel_args* args);
     void ExitIdleLocked();
     void ResetBackoffLocked();
@@ -176,7 +197,6 @@ class XdsClusterManagerLb : public LoadBalancingPolicy {
 
     RefCountedPtr<ChildPickerWrapper> picker_wrapper_;
     grpc_connectivity_state connectivity_state_ = GRPC_CHANNEL_IDLE;
-    bool seen_failure_since_ready_ = false;
 
     // States for delayed removal.
     grpc_timer delayed_removal_timer_;
@@ -196,6 +216,7 @@ class XdsClusterManagerLb : public LoadBalancingPolicy {
 
   // Internal state.
   bool shutting_down_ = false;
+  bool update_in_progress_ = false;
 
   // Children.
   std::map<std::string, OrphanablePtr<ClusterChild>> children_;
@@ -207,8 +228,10 @@ class XdsClusterManagerLb : public LoadBalancingPolicy {
 
 XdsClusterManagerLb::PickResult XdsClusterManagerLb::ClusterPicker::Pick(
     PickArgs args) {
+  auto* call_state = static_cast<ClientChannel::LoadBalancedCall::LbCallState*>(
+      args.call_state);
   auto cluster_name =
-      args.call_state->ExperimentalGetCallAttribute(kXdsClusterAttribute);
+      call_state->GetCallAttribute(XdsClusterAttributeTypeName());
   auto it = cluster_map_.find(cluster_name);
   if (it != cluster_map_.end()) {
     return it->second->Pick(args);
@@ -254,6 +277,7 @@ void XdsClusterManagerLb::UpdateLocked(UpdateArgs args) {
   if (GRPC_TRACE_FLAG_ENABLED(grpc_xds_cluster_manager_lb_trace)) {
     gpr_log(GPR_INFO, "[xds_cluster_manager_lb %p] Received update", this);
   }
+  update_in_progress_ = true;
   // Update config.
   config_ = std::move(args.config);
   // Deactivate the children not in the new config.
@@ -268,19 +292,24 @@ void XdsClusterManagerLb::UpdateLocked(UpdateArgs args) {
   for (const auto& p : config_->cluster_map()) {
     const std::string& name = p.first;
     const RefCountedPtr<LoadBalancingPolicy::Config>& config = p.second;
-    auto it = children_.find(name);
-    if (it == children_.end()) {
-      it = children_
-               .emplace(name, MakeOrphanable<ClusterChild>(
-                                  Ref(DEBUG_LOCATION, "ClusterChild"), name))
-               .first;
+    auto& child = children_[name];
+    if (child == nullptr) {
+      child = MakeOrphanable<ClusterChild>(Ref(DEBUG_LOCATION, "ClusterChild"),
+                                           name);
     }
-    it->second->UpdateLocked(config, args.addresses, args.args);
+    child->UpdateLocked(config, args.addresses, args.args);
   }
+  update_in_progress_ = false;
   UpdateStateLocked();
 }
 
 void XdsClusterManagerLb::UpdateStateLocked() {
+  // If we're in the process of propagating an update from our parent to
+  // our children, ignore any updates that come from the children.  We
+  // will instead return a new picker once the update has been seen by
+  // all children.  This avoids unnecessary picker churn while an update
+  // is being propagated to our children.
+  if (update_in_progress_) return;
   // Also count the number of children in each state, to determine the
   // overall state.
   size_t num_ready = 0;
@@ -441,7 +470,8 @@ XdsClusterManagerLb::ClusterChild::CreateChildPolicyLocked(
 
 void XdsClusterManagerLb::ClusterChild::UpdateLocked(
     RefCountedPtr<LoadBalancingPolicy::Config> config,
-    const ServerAddressList& addresses, const grpc_channel_args* args) {
+    const absl::StatusOr<ServerAddressList>& addresses,
+    const grpc_channel_args* args) {
   if (xds_cluster_manager_policy_->shutting_down_) return;
   // Update child weight.
   // Reactivate if needed.
@@ -486,7 +516,8 @@ void XdsClusterManagerLb::ClusterChild::DeactivateLocked() {
   Ref(DEBUG_LOCATION, "ClusterChild+timer").release();
   grpc_timer_init(&delayed_removal_timer_,
                   ExecCtx::Get()->Now() +
-                      GRPC_XDS_CLUSTER_MANAGER_CHILD_RETENTION_INTERVAL_MS,
+                      Duration::Milliseconds(
+                          GRPC_XDS_CLUSTER_MANAGER_CHILD_RETENTION_INTERVAL_MS),
                   &on_delayed_removal_timer_);
   delayed_removal_timer_callback_pending_ = true;
 }
@@ -503,7 +534,7 @@ void XdsClusterManagerLb::ClusterChild::OnDelayedRemovalTimer(
 void XdsClusterManagerLb::ClusterChild::OnDelayedRemovalTimerLocked(
     grpc_error_handle error) {
   delayed_removal_timer_callback_pending_ = false;
-  if (error == GRPC_ERROR_NONE && !shutdown_) {
+  if (GRPC_ERROR_IS_NONE(error) && !shutdown_) {
     xds_cluster_manager_policy_->children_.erase(name_);
   }
   Unref(DEBUG_LOCATION, "ClusterChild+timer");
@@ -545,19 +576,13 @@ void XdsClusterManagerLb::ClusterChild::Helper::UpdateState(
       MakeRefCounted<ChildPickerWrapper>(xds_cluster_manager_child_->name_,
                                          std::move(picker));
   // Decide what state to report for aggregation purposes.
-  // If we haven't seen a failure since the last time we were in state
-  // READY, then we report the state change as-is.  However, once we do see
-  // a failure, we report TRANSIENT_FAILURE and ignore any subsequent state
-  // changes until we go back into state READY.
-  if (!xds_cluster_manager_child_->seen_failure_since_ready_) {
-    if (state == GRPC_CHANNEL_TRANSIENT_FAILURE) {
-      xds_cluster_manager_child_->seen_failure_since_ready_ = true;
-    }
-  } else {
-    if (state != GRPC_CHANNEL_READY) return;
-    xds_cluster_manager_child_->seen_failure_since_ready_ = false;
+  // If the last recorded state was TRANSIENT_FAILURE and the new state
+  // is something other than READY, don't change the state.
+  if (xds_cluster_manager_child_->connectivity_state_ !=
+          GRPC_CHANNEL_TRANSIENT_FAILURE ||
+      state == GRPC_CHANNEL_READY) {
+    xds_cluster_manager_child_->connectivity_state_ = state;
   }
-  xds_cluster_manager_child_->connectivity_state_ = state;
   // Notify the LB policy.
   xds_cluster_manager_child_->xds_cluster_manager_policy_->UpdateStateLocked();
 }
@@ -602,7 +627,7 @@ class XdsClusterManagerLbFactory : public LoadBalancingPolicyFactory {
 
   RefCountedPtr<LoadBalancingPolicy::Config> ParseLoadBalancingConfig(
       const Json& json, grpc_error_handle* error) const override {
-    GPR_DEBUG_ASSERT(error != nullptr && *error == GRPC_ERROR_NONE);
+    GPR_DEBUG_ASSERT(error != nullptr && GRPC_ERROR_IS_NONE(*error));
     if (json.type() == Json::Type::JSON_NULL) {
       // xds_cluster_manager was mentioned as a policy in the deprecated
       // loadBalancingPolicy field or in the client API.
@@ -673,7 +698,7 @@ class XdsClusterManagerLbFactory : public LoadBalancingPolicyFactory {
       *child_config = LoadBalancingPolicyRegistry::ParseLoadBalancingConfig(
           it->second, &parse_error);
       if (*child_config == nullptr) {
-        GPR_DEBUG_ASSERT(parse_error != GRPC_ERROR_NONE);
+        GPR_DEBUG_ASSERT(!GRPC_ERROR_IS_NONE(parse_error));
         std::vector<grpc_error_handle> child_errors;
         child_errors.push_back(parse_error);
         error_list.push_back(