grpc 1.64.3 → 1.71.0

data/src/core/load_balancing/pick_first/pick_first.cc

@@ -16,10 +16,15 @@
 
 #include "src/core/load_balancing/pick_first/pick_first.h"
 
+#include <grpc/event_engine/event_engine.h>
+#include <grpc/impl/channel_arg_names.h>
+#include <grpc/impl/connectivity_state.h>
+#include <grpc/support/port_platform.h>
 #include <inttypes.h>
 #include <string.h>
 
 #include <memory>
+#include <optional>
 #include <set>
 #include <string>
 #include <type_traits>
@@ -28,49 +33,40 @@
 
 #include "absl/algorithm/container.h"
 #include "absl/log/check.h"
+#include "absl/log/log.h"
 #include "absl/random/random.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 <grpc/event_engine/event_engine.h>
-#include <grpc/impl/channel_arg_names.h>
-#include <grpc/impl/connectivity_state.h>
-#include <grpc/support/log.h>
-#include <grpc/support/port_platform.h>
-
+#include "src/core/config/core_configuration.h"
 #include "src/core/lib/address_utils/sockaddr_utils.h"
 #include "src/core/lib/channel/channel_args.h"
-#include "src/core/lib/channel/metrics.h"
-#include "src/core/lib/config/core_configuration.h"
 #include "src/core/lib/debug/trace.h"
 #include "src/core/lib/experiments/experiments.h"
-#include "src/core/lib/gpr/useful.h"
-#include "src/core/lib/gprpp/crash.h"
-#include "src/core/lib/gprpp/debug_location.h"
-#include "src/core/lib/gprpp/orphanable.h"
-#include "src/core/lib/gprpp/ref_counted_ptr.h"
-#include "src/core/lib/gprpp/time.h"
-#include "src/core/lib/gprpp/work_serializer.h"
 #include "src/core/lib/iomgr/exec_ctx.h"
 #include "src/core/lib/iomgr/iomgr_fwd.h"
 #include "src/core/lib/iomgr/resolved_address.h"
-#include "src/core/lib/json/json.h"
-#include "src/core/lib/json/json_args.h"
-#include "src/core/lib/json/json_object_loader.h"
 #include "src/core/lib/transport/connectivity_state.h"
 #include "src/core/load_balancing/health_check_client.h"
 #include "src/core/load_balancing/lb_policy.h"
 #include "src/core/load_balancing/lb_policy_factory.h"
 #include "src/core/load_balancing/subchannel_interface.h"
 #include "src/core/resolver/endpoint_addresses.h"
+#include "src/core/telemetry/metrics.h"
+#include "src/core/util/crash.h"
+#include "src/core/util/debug_location.h"
+#include "src/core/util/json/json.h"
+#include "src/core/util/json/json_args.h"
+#include "src/core/util/json/json_object_loader.h"
+#include "src/core/util/orphanable.h"
+#include "src/core/util/ref_counted_ptr.h"
+#include "src/core/util/time.h"
+#include "src/core/util/useful.h"
+#include "src/core/util/work_serializer.h"
 
 namespace grpc_core {
 
-TraceFlag grpc_lb_pick_first_trace(false, "pick_first");
-
 namespace {
 
 //
@@ -84,19 +80,25 @@ const auto kMetricDisconnections =
         "grpc.lb.pick_first.disconnections",
         "EXPERIMENTAL.  Number of times the selected subchannel becomes "
         "disconnected.",
-        "{disconnection}", {kMetricLabelTarget}, {}, false);
+        "{disconnection}", false)
+        .Labels(kMetricLabelTarget)
+        .Build();
 
 const auto kMetricConnectionAttemptsSucceeded =
     GlobalInstrumentsRegistry::RegisterUInt64Counter(
         "grpc.lb.pick_first.connection_attempts_succeeded",
         "EXPERIMENTAL.  Number of successful connection attempts.", "{attempt}",
-        {kMetricLabelTarget}, {}, false);
+        false)
+        .Labels(kMetricLabelTarget)
+        .Build();
 
 const auto kMetricConnectionAttemptsFailed =
     GlobalInstrumentsRegistry::RegisterUInt64Counter(
         "grpc.lb.pick_first.connection_attempts_failed",
         "EXPERIMENTAL.  Number of failed connection attempts.", "{attempt}",
-        {kMetricLabelTarget}, {}, false);
+        false)
+        .Labels(kMetricLabelTarget)
+        .Build();
 
 class PickFirstConfig final : public LoadBalancingPolicy::Config {
  public:
@@ -193,7 +195,7 @@ class PickFirst final : public LoadBalancingPolicy {
 
         // TODO(roth): Once we remove pollset_set, we should no longer
         // need to hold a ref to PickFirst.  Instead, we can make this a
-        // raw pointer and put it in an absl::variant with subchannel_data_.
+        // raw pointer and put it in an std::variant with subchannel_data_.
         RefCountedPtr<PickFirst> pick_first_;
 
         RefCountedPtr<SubchannelInterface> subchannel_;
@@ -204,7 +206,7 @@ class PickFirst final : public LoadBalancingPolicy {
       SubchannelData(SubchannelList* subchannel_list, size_t index,
                      RefCountedPtr<SubchannelInterface> subchannel);
 
-      absl::optional<grpc_connectivity_state> connectivity_state() const {
+      std::optional<grpc_connectivity_state> connectivity_state() const {
         return connectivity_state_;
       }
       const absl::Status& connectivity_status() const {
@@ -222,6 +224,7 @@ class PickFirst final : public LoadBalancingPolicy {
       void RequestConnectionWithTimer();
 
       bool seen_transient_failure() const { return seen_transient_failure_; }
+      void set_seen_transient_failure() { seen_transient_failure_ = true; }
 
      private:
       // This method will be invoked once soon after instantiation to report
@@ -237,14 +240,14 @@ class PickFirst final : public LoadBalancingPolicy {
       // Subchannel state.
       OrphanablePtr<SubchannelState> subchannel_state_;
       // Data updated by the watcher.
-      absl::optional<grpc_connectivity_state> connectivity_state_;
+      std::optional<grpc_connectivity_state> connectivity_state_;
       absl::Status connectivity_status_;
       bool seen_transient_failure_ = false;
     };
 
     SubchannelList(RefCountedPtr<PickFirst> policy,
                    EndpointAddressesIterator* addresses,
-                   const ChannelArgs& args);
+                   const ChannelArgs& args, absl::string_view resolution_note);
 
     ~SubchannelList() override;
 
@@ -267,6 +270,8 @@ class PickFirst final : public LoadBalancingPolicy {
       return true;
     }
 
+    void ReportTransientFailure(absl::Status status);
+
    private:
     // Returns true if all subchannels have seen their initial
     // connectivity state notifications.
@@ -291,6 +296,7 @@ class PickFirst final : public LoadBalancingPolicy {
     RefCountedPtr<PickFirst> policy_;
 
     ChannelArgs args_;
+    std::string resolution_note_;
 
     // The list of subchannels.
     std::vector<std::unique_ptr<SubchannelData>> subchannels_;
@@ -307,7 +313,7 @@ class PickFirst final : public LoadBalancingPolicy {
     // initial pass is over, this will be equal to size().
     size_t attempting_index_ = 0;
     // Happy Eyeballs timer handle.
-    absl::optional<grpc_event_engine::experimental::EventEngine::TaskHandle>
+    std::optional<grpc_event_engine::experimental::EventEngine::TaskHandle>
         timer_handle_;
 
     // After the initial Happy Eyeballs pass, the number of failures
@@ -321,8 +327,9 @@ class PickFirst final : public LoadBalancingPolicy {
   class HealthWatcher final
       : public SubchannelInterface::ConnectivityStateWatcherInterface {
    public:
-    explicit HealthWatcher(RefCountedPtr<PickFirst> policy)
-        : policy_(std::move(policy)) {}
+    HealthWatcher(RefCountedPtr<PickFirst> policy,
+                  absl::string_view resolution_note)
+        : policy_(std::move(policy)), resolution_note_(resolution_note) {}
 
     ~HealthWatcher() override {
       policy_.reset(DEBUG_LOCATION, "HealthWatcher dtor");
@@ -337,6 +344,7 @@ class PickFirst final : public LoadBalancingPolicy {
 
    private:
     RefCountedPtr<PickFirst> policy_;
+    std::string resolution_note_;
   };
 
   class Picker final : public SubchannelPicker {
@@ -416,22 +424,16 @@ PickFirst::PickFirst(Args args)
                     .GetInt(GRPC_ARG_HAPPY_EYEBALLS_CONNECTION_ATTEMPT_DELAY_MS)
                     .value_or(250),
                 100, 2000))) {
-  if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-    gpr_log(GPR_INFO, "Pick First %p created.", this);
-  }
+  GRPC_TRACE_LOG(pick_first, INFO) << "Pick First " << this << " created.";
 }
 
 PickFirst::~PickFirst() {
-  if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-    gpr_log(GPR_INFO, "Destroying Pick First %p", this);
-  }
-  CHECK(subchannel_list_ == nullptr);
+  GRPC_TRACE_LOG(pick_first, INFO) << "Destroying Pick First " << this;
+  CHECK_EQ(subchannel_list_.get(), nullptr);
 }
 
 void PickFirst::ShutdownLocked() {
-  if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-    gpr_log(GPR_INFO, "Pick First %p Shutting down", this);
-  }
+  GRPC_TRACE_LOG(pick_first, INFO) << "Pick First " << this << " Shutting down";
   shutdown_ = true;
   UnsetSelectedSubchannel();
   subchannel_list_.reset();
@@ -440,9 +442,8 @@ void PickFirst::ShutdownLocked() {
 void PickFirst::ExitIdleLocked() {
   if (shutdown_) return;
   if (IsIdle()) {
-    if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-      gpr_log(GPR_INFO, "Pick First %p exiting idle", this);
-    }
+    GRPC_TRACE_LOG(pick_first, INFO)
+        << "Pick First " << this << " exiting idle";
     AttemptToConnectUsingLatestUpdateArgsLocked();
   }
 }
@@ -458,26 +459,22 @@ void PickFirst::AttemptToConnectUsingLatestUpdateArgsLocked() {
     addresses = latest_update_args_.addresses->get();
   }
   // Replace subchannel_list_.
-  if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace) &&
-      subchannel_list_ != nullptr) {
-    gpr_log(GPR_INFO, "[PF %p] Shutting down previous subchannel list %p", this,
-            subchannel_list_.get());
+  if (GRPC_TRACE_FLAG_ENABLED(pick_first) && subchannel_list_ != nullptr) {
+    LOG(INFO) << "[PF " << this << "] Shutting down previous subchannel list "
+              << subchannel_list_.get();
   }
   subchannel_list_ = MakeOrphanable<SubchannelList>(
       RefAsSubclass<PickFirst>(DEBUG_LOCATION, "SubchannelList"), addresses,
-      latest_update_args_.args);
+      latest_update_args_.args, latest_update_args_.resolution_note);
   // Empty update or no valid subchannels.  Put the channel in
   // TRANSIENT_FAILURE and request re-resolution.  Also unset the
   // current selected subchannel.
   if (subchannel_list_->size() == 0) {
     channel_control_helper()->RequestReresolution();
-    absl::Status status =
-        latest_update_args_.addresses.ok()
-            ? absl::UnavailableError(absl::StrCat(
-                  "empty address list: ", latest_update_args_.resolution_note))
-            : latest_update_args_.addresses.status();
-    UpdateState(GRPC_CHANNEL_TRANSIENT_FAILURE, status,
-                MakeRefCounted<TransientFailurePicker>(status));
+    absl::Status status = latest_update_args_.addresses.ok()
+                              ? absl::UnavailableError("empty address list")
+                              : latest_update_args_.addresses.status();
+    subchannel_list_->ReportTransientFailure(std::move(status));
     UnsetSelectedSubchannel();
   }
 }
@@ -512,12 +509,13 @@ class AddressFamilyIterator final {
 };
 
 absl::Status PickFirst::UpdateLocked(UpdateArgs args) {
-  if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
+  if (GRPC_TRACE_FLAG_ENABLED(pick_first)) {
     if (args.addresses.ok()) {
-      gpr_log(GPR_INFO, "Pick First %p received update", this);
+      LOG(INFO) << "Pick First " << this << " received update";
     } else {
-      gpr_log(GPR_INFO, "Pick First %p received update with address error: %s",
-              this, args.addresses.status().ToString().c_str());
+      LOG(INFO) << "Pick First " << this
+                << " received update with address error: "
+                << args.addresses.status();
     }
   }
   // Set return status based on the address list.
@@ -616,7 +614,7 @@ void PickFirst::GoIdle() {
   // ExitIdleLocked() instead.
   channel_control_helper()->RequestReresolution();
   // Enter idle.
-  UpdateState(GRPC_CHANNEL_IDLE, absl::Status(),
+  UpdateState(GRPC_CHANNEL_IDLE, absl::OkStatus(),
               MakeRefCounted<QueuePicker>(Ref(DEBUG_LOCATION, "QueuePicker")));
 }
 
@@ -627,11 +625,10 @@ void PickFirst::GoIdle() {
 void PickFirst::HealthWatcher::OnConnectivityStateChange(
     grpc_connectivity_state new_state, absl::Status status) {
   if (policy_->health_watcher_ != this) return;
-  if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-    gpr_log(GPR_INFO, "[PF %p] health watch state update: %s (%s)",
-            policy_.get(), ConnectivityStateName(new_state),
-            status.ToString().c_str());
-  }
+  GRPC_TRACE_LOG(pick_first, INFO)
+      << "[PF " << policy_.get()
+      << "] health watch state update: " << ConnectivityStateName(new_state)
+      << " (" << status << ")";
   switch (new_state) {
     case GRPC_CHANNEL_READY:
       policy_->channel_control_helper()->UpdateState(
@@ -649,11 +646,17 @@ void PickFirst::HealthWatcher::OnConnectivityStateChange(
           new_state, absl::OkStatus(),
           MakeRefCounted<QueuePicker>(policy_->Ref()));
       break;
-    case GRPC_CHANNEL_TRANSIENT_FAILURE:
+    case GRPC_CHANNEL_TRANSIENT_FAILURE: {
+      std::string message = absl::StrCat("health watch: ", status.message());
+      if (!resolution_note_.empty()) {
+        absl::StrAppend(&message, " (", resolution_note_, ")");
+      }
       policy_->channel_control_helper()->UpdateState(
           GRPC_CHANNEL_TRANSIENT_FAILURE, status,
-          MakeRefCounted<TransientFailurePicker>(status));
+          MakeRefCounted<TransientFailurePicker>(
+              absl::UnavailableError(message)));
       break;
+    }
     case GRPC_CHANNEL_SHUTDOWN:
       Crash("health watcher reported state SHUTDOWN");
   }
@@ -669,23 +672,19 @@ PickFirst::SubchannelList::SubchannelData::SubchannelState::SubchannelState(
     : subchannel_data_(subchannel_data),
       pick_first_(subchannel_data_->subchannel_list_->policy_),
       subchannel_(std::move(subchannel)) {
-  if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-    gpr_log(GPR_INFO,
-            "[PF %p] subchannel state %p (subchannel %p): starting watch",
-            pick_first_.get(), this, subchannel_.get());
-  }
+  GRPC_TRACE_LOG(pick_first, INFO)
+      << "[PF " << pick_first_.get() << "] subchannel state " << this
+      << " (subchannel " << subchannel_.get() << "): starting watch";
   auto watcher = std::make_unique<Watcher>(Ref(DEBUG_LOCATION, "Watcher"));
   watcher_ = watcher.get();
   subchannel_->WatchConnectivityState(std::move(watcher));
 }
 
 void PickFirst::SubchannelList::SubchannelData::SubchannelState::Orphan() {
-  if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-    gpr_log(GPR_INFO,
-            "[PF %p] subchannel state %p (subchannel %p): "
-            "cancelling watch and unreffing subchannel",
-            pick_first_.get(), this, subchannel_.get());
-  }
+  GRPC_TRACE_LOG(pick_first, INFO)
+      << "[PF " << pick_first_.get() << "] subchannel state " << this
+      << " (subchannel " << subchannel_.get()
+      << "): cancelling watch and unreffing subchannel";
   subchannel_data_ = nullptr;
   subchannel_->CancelConnectivityStateWatch(watcher_);
   watcher_ = nullptr;
@@ -695,10 +694,9 @@ void PickFirst::SubchannelList::SubchannelData::SubchannelState::Orphan() {
 }
 
 void PickFirst::SubchannelList::SubchannelData::SubchannelState::Select() {
-  if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-    gpr_log(GPR_INFO, "Pick First %p selected subchannel %p", pick_first_.get(),
-            subchannel_.get());
-  }
+  GRPC_TRACE_LOG(pick_first, INFO)
+      << "Pick First " << pick_first_.get() << " selected subchannel "
+      << subchannel_.get();
   CHECK_NE(subchannel_data_, nullptr);
   pick_first_->UnsetSelectedSubchannel();  // Cancel health watch, if any.
   pick_first_->selected_ = std::move(subchannel_data_->subchannel_state_);
@@ -707,11 +705,11 @@ void PickFirst::SubchannelList::SubchannelData::SubchannelState::Select() {
   // for the health status notification.
   // If health checking is NOT enabled, report READY.
   if (pick_first_->enable_health_watch_) {
-    if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-      gpr_log(GPR_INFO, "[PF %p] starting health watch", pick_first_.get());
-    }
+    GRPC_TRACE_LOG(pick_first, INFO)
+        << "[PF " << pick_first_.get() << "] starting health watch";
     auto watcher = std::make_unique<HealthWatcher>(
-        pick_first_.Ref(DEBUG_LOCATION, "HealthWatcher"));
+        pick_first_.Ref(DEBUG_LOCATION, "HealthWatcher"),
+        subchannel_data_->subchannel_list_->resolution_note_);
     pick_first_->health_watcher_ = watcher.get();
     auto health_data_watcher = MakeHealthCheckWatcher(
         pick_first_->work_serializer(),
@@ -719,7 +717,7 @@ void PickFirst::SubchannelList::SubchannelData::SubchannelState::Select() {
     pick_first_->health_data_watcher_ = health_data_watcher.get();
     subchannel_->AddDataWatcher(std::move(health_data_watcher));
   } else {
-    pick_first_->UpdateState(GRPC_CHANNEL_READY, absl::Status(),
+    pick_first_->UpdateState(GRPC_CHANNEL_READY, absl::OkStatus(),
                              MakeRefCounted<Picker>(subchannel_));
   }
   // Report successful connection.
@@ -745,18 +743,18 @@ void PickFirst::SubchannelList::SubchannelData::SubchannelState::
     OnConnectivityStateChange(grpc_connectivity_state new_state,
                               absl::Status status) {
   if (watcher_ == nullptr) return;
-  if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-    gpr_log(GPR_INFO,
-            "[PF %p] subchannel state %p (subchannel %p): connectivity "
-            "changed: new_state=%s, status=%s, watcher=%p, "
-            "subchannel_data_=%p, pick_first_->selected_=%p",
-            pick_first_.get(), this, subchannel_.get(),
-            ConnectivityStateName(new_state), status.ToString().c_str(),
-            watcher_, subchannel_data_, pick_first_->selected_.get());
-  }
+  GRPC_TRACE_LOG(pick_first, INFO)
+      << "[PF " << pick_first_.get() << "] subchannel state " << this
+      << " (subchannel " << subchannel_.get()
+      << "): connectivity changed: new_state="
+      << ConnectivityStateName(new_state) << ", status=" << status
+      << ", watcher=" << watcher_ << ", subchannel_data_=" << subchannel_data_
+      << ", pick_first_->selected_=" << pick_first_->selected_.get();
   // If we're still part of a subchannel list trying to connect, check
   // if we're connected.
   if (subchannel_data_ != nullptr) {
+    CHECK_EQ(pick_first_->subchannel_list_.get(),
+             subchannel_data_->subchannel_list_);
     // If the subchannel is READY, use it.
     // Otherwise, tell the subchannel list to keep trying.
     if (new_state == GRPC_CHANNEL_READY) {
@@ -767,12 +765,11 @@ void PickFirst::SubchannelList::SubchannelData::SubchannelState::
     return;
   }
   // We aren't trying to connect, so we must be the selected subchannel.
-  CHECK(pick_first_->selected_.get() == this);
-  if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-    gpr_log(GPR_INFO,
-            "Pick First %p selected subchannel connectivity changed to %s",
-            pick_first_.get(), ConnectivityStateName(new_state));
-  }
+  CHECK_EQ(pick_first_->selected_.get(), this);
+  GRPC_TRACE_LOG(pick_first, INFO)
+      << "Pick First " << pick_first_.get()
+      << " selected subchannel connectivity changed to "
+      << ConnectivityStateName(new_state);
   // Any state change is considered to be a failure of the existing
   // connection.  Report the failure.
   auto& stats_plugins =
@@ -792,12 +789,10 @@ PickFirst::SubchannelList::SubchannelData::SubchannelData(
     SubchannelList* subchannel_list, size_t index,
     RefCountedPtr<SubchannelInterface> subchannel)
     : subchannel_list_(subchannel_list), index_(index) {
-  if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-    gpr_log(GPR_INFO,
-            "[PF %p] subchannel list %p index %" PRIuPTR
-            ": creating subchannel data",
-            subchannel_list_->policy_.get(), subchannel_list_, index_);
-  }
+  GRPC_TRACE_LOG(pick_first, INFO)
+      << "[PF " << subchannel_list_->policy_.get() << "] subchannel list "
+      << subchannel_list_ << " index " << index_
+      << ": creating subchannel data";
   subchannel_state_ =
       MakeOrphanable<SubchannelState>(this, std::move(subchannel));
 }
@@ -805,38 +800,34 @@ PickFirst::SubchannelList::SubchannelData::SubchannelData(
 void PickFirst::SubchannelList::SubchannelData::OnConnectivityStateChange(
     grpc_connectivity_state new_state, absl::Status status) {
   PickFirst* p = subchannel_list_->policy_.get();
-  if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-    gpr_log(
-        GPR_INFO,
-        "[PF %p] subchannel list %p index %" PRIuPTR " of %" PRIuPTR
-        " (subchannel_state %p): connectivity changed: old_state=%s, "
-        "new_state=%s, status=%s, seen_transient_failure=%d, p->selected_=%p, "
-        "p->subchannel_list_=%p, p->subchannel_list_->shutting_down_=%d",
-        p, subchannel_list_, index_, subchannel_list_->size(),
-        subchannel_state_.get(),
-        (connectivity_state_.has_value()
-             ? ConnectivityStateName(*connectivity_state_)
-             : "N/A"),
-        ConnectivityStateName(new_state), status.ToString().c_str(),
-        seen_transient_failure_, p->selected_.get(), p->subchannel_list_.get(),
-        p->subchannel_list_->shutting_down_);
-  }
+  GRPC_TRACE_LOG(pick_first, INFO)
+      << "[PF " << p << "] subchannel list " << subchannel_list_ << " index "
+      << index_ << " of " << subchannel_list_->size() << " (subchannel_state "
+      << subchannel_state_.get() << "): connectivity changed: old_state="
+      << (connectivity_state_.has_value()
+              ? ConnectivityStateName(*connectivity_state_)
+              : "N/A")
+      << ", new_state=" << ConnectivityStateName(new_state)
+      << ", status=" << status
+      << ", seen_transient_failure=" << seen_transient_failure_
+      << ", p->selected_=" << p->selected_.get()
+      << ", p->subchannel_list_=" << p->subchannel_list_.get()
+      << ", p->subchannel_list_->shutting_down_="
+      << p->subchannel_list_->shutting_down_;
   if (subchannel_list_->shutting_down_) return;
   // The notification must be for a subchannel in the current list.
-  CHECK(subchannel_list_ == p->subchannel_list_.get());
+  CHECK_EQ(subchannel_list_, p->subchannel_list_.get());
   // SHUTDOWN should never happen.
-  CHECK(new_state != GRPC_CHANNEL_SHUTDOWN);
+  CHECK_NE(new_state, GRPC_CHANNEL_SHUTDOWN);
   // READY should be caught by SubchannelState, in which case it will
   // not call us in the first place.
-  CHECK(new_state != GRPC_CHANNEL_READY);
+  CHECK_NE(new_state, GRPC_CHANNEL_READY);
   // Update state.
-  absl::optional<grpc_connectivity_state> old_state = connectivity_state_;
+  std::optional<grpc_connectivity_state> old_state = connectivity_state_;
   connectivity_state_ = new_state;
   connectivity_status_ = std::move(status);
   // Make sure we note when a subchannel has seen TRANSIENT_FAILURE.
-  bool prev_seen_transient_failure = seen_transient_failure_;
   if (new_state == GRPC_CHANNEL_TRANSIENT_FAILURE) {
-    seen_transient_failure_ = true;
     subchannel_list_->last_failure_ = connectivity_status_;
   }
   // If this is the initial connectivity state update for this subchannel,
@@ -858,12 +849,10 @@ void PickFirst::SubchannelList::SubchannelData::OnConnectivityStateChange(
     // is not in the new list.  In that case, we drop the current
     // connection and report IDLE.
     if (p->selected_ != nullptr) {
-      if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-        gpr_log(GPR_INFO,
-                "[PF %p] subchannel list %p: new update has no subchannels in "
-                "state READY; dropping existing connection and going IDLE",
-                p, subchannel_list_);
-      }
+      GRPC_TRACE_LOG(pick_first, INFO)
+          << "[PF " << p << "] subchannel list " << subchannel_list_
+          << ": new update has no subchannels in state READY; dropping "
+             "existing connection and going IDLE";
       p->GoIdle();
     } else {
       // Start trying to connect, starting with the first subchannel.
@@ -883,6 +872,8 @@ void PickFirst::SubchannelList::SubchannelData::OnConnectivityStateChange(
   // Otherwise, process connectivity state change.
   switch (*connectivity_state_) {
     case GRPC_CHANNEL_TRANSIENT_FAILURE: {
+      bool prev_seen_transient_failure =
+          std::exchange(seen_transient_failure_, true);
       // If this is the first failure we've seen on this subchannel,
       // then we're still in the Happy Eyeballs pass.
       if (!prev_seen_transient_failure && seen_transient_failure_) {
@@ -920,8 +911,7 @@ void PickFirst::SubchannelList::SubchannelData::OnConnectivityStateChange(
                    ? ""
                    : "failed to connect to all addresses; last error: "),
               connectivity_status_.ToString()));
-          p->UpdateState(GRPC_CHANNEL_TRANSIENT_FAILURE, status,
-                         MakeRefCounted<TransientFailurePicker>(status));
+          subchannel_list_->ReportTransientFailure(std::move(status));
         }
       }
       break;
@@ -939,7 +929,7 @@ void PickFirst::SubchannelList::SubchannelData::OnConnectivityStateChange(
       // TRANSIENT_FAILURE.
       // TODO(roth): Squelch duplicate CONNECTING updates.
       if (p->state_ != GRPC_CHANNEL_TRANSIENT_FAILURE) {
-        p->UpdateState(GRPC_CHANNEL_CONNECTING, absl::Status(),
+        p->UpdateState(GRPC_CHANNEL_CONNECTING, absl::OkStatus(),
                        MakeRefCounted<QueuePicker>(nullptr));
       }
       break;
@@ -954,44 +944,36 @@ void PickFirst::SubchannelList::SubchannelData::RequestConnectionWithTimer() {
   if (connectivity_state_ == GRPC_CHANNEL_IDLE) {
     subchannel_state_->RequestConnection();
   } else {
-    CHECK(connectivity_state_ == GRPC_CHANNEL_CONNECTING);
+    CHECK_EQ(connectivity_state_.value(), GRPC_CHANNEL_CONNECTING);
   }
   // If this is not the last subchannel in the list, start the timer.
   if (index_ != subchannel_list_->size() - 1) {
     PickFirst* p = subchannel_list_->policy_.get();
-    if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-      gpr_log(GPR_INFO,
-              "Pick First %p subchannel list %p: starting Connection "
-              "Attempt Delay timer for %" PRId64 "ms for index %" PRIuPTR,
-              p, subchannel_list_, p->connection_attempt_delay_.millis(),
-              index_);
-    }
+    GRPC_TRACE_LOG(pick_first, INFO)
+        << "Pick First " << p << " subchannel list " << subchannel_list_
+        << ": starting Connection Attempt Delay timer for "
+        << p->connection_attempt_delay_.millis() << "ms for index " << index_;
     subchannel_list_->timer_handle_ =
         p->channel_control_helper()->GetEventEngine()->RunAfter(
             p->connection_attempt_delay_,
             [subchannel_list =
                  subchannel_list_->Ref(DEBUG_LOCATION, "timer")]() mutable {
-              ApplicationCallbackExecCtx application_exec_ctx;
               ExecCtx exec_ctx;
               auto* sl = subchannel_list.get();
               sl->policy_->work_serializer()->Run(
                   [subchannel_list = std::move(subchannel_list)]() {
-                    if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-                      gpr_log(GPR_INFO,
-                              "Pick First %p subchannel list %p: Connection "
-                              "Attempt Delay timer fired (shutting_down=%d, "
-                              "selected=%p)",
-                              subchannel_list->policy_.get(),
-                              subchannel_list.get(),
-                              subchannel_list->shutting_down_,
-                              subchannel_list->policy_->selected_.get());
-                    }
+                    GRPC_TRACE_LOG(pick_first, INFO)
+                        << "Pick First " << subchannel_list->policy_.get()
+                        << " subchannel list " << subchannel_list.get()
+                        << ": Connection Attempt Delay timer fired "
+                           "(shutting_down="
+                        << subchannel_list->shutting_down_ << ", selected="
+                        << subchannel_list->policy_->selected_.get() << ")";
                     if (subchannel_list->shutting_down_) return;
                     if (subchannel_list->policy_->selected_ != nullptr) return;
                     ++subchannel_list->attempting_index_;
                     subchannel_list->StartConnectingNextSubchannel();
-                  },
-                  DEBUG_LOCATION);
+                  });
             });
   }
 }
@@ -1002,18 +984,18 @@ void PickFirst::SubchannelList::SubchannelData::RequestConnectionWithTimer() {
 
 PickFirst::SubchannelList::SubchannelList(RefCountedPtr<PickFirst> policy,
                                           EndpointAddressesIterator* addresses,
-                                          const ChannelArgs& args)
+                                          const ChannelArgs& args,
+                                          absl::string_view resolution_note)
     : InternallyRefCounted<SubchannelList>(
-          GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace) ? "SubchannelList"
-                                                            : nullptr),
+          GRPC_TRACE_FLAG_ENABLED(pick_first) ? "SubchannelList" : nullptr),
       policy_(std::move(policy)),
-      args_(args.Remove(GRPC_ARG_INTERNAL_PICK_FIRST_ENABLE_HEALTH_CHECKING)
-                .Remove(
-                    GRPC_ARG_INTERNAL_PICK_FIRST_OMIT_STATUS_MESSAGE_PREFIX)) {
-  if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-    gpr_log(GPR_INFO, "[PF %p] Creating subchannel list %p - channel args: %s",
-            policy_.get(), this, args_.ToString().c_str());
-  }
+      args_(
+          args.Remove(GRPC_ARG_INTERNAL_PICK_FIRST_ENABLE_HEALTH_CHECKING)
+              .Remove(GRPC_ARG_INTERNAL_PICK_FIRST_OMIT_STATUS_MESSAGE_PREFIX)),
+      resolution_note_(resolution_note) {
+  GRPC_TRACE_LOG(pick_first, INFO)
+      << "[PF " << policy_.get() << "] Creating subchannel list " << this
+      << " - channel args: " << args_.ToString();
   if (addresses == nullptr) return;
   // Create a subchannel for each address.
   addresses->ForEach([&](const EndpointAddresses& address) {
@@ -1023,39 +1005,33 @@ PickFirst::SubchannelList::SubchannelList(RefCountedPtr<PickFirst> policy,
             address.address(), address.args(), args_);
     if (subchannel == nullptr) {
       // Subchannel could not be created.
-      if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-        gpr_log(GPR_INFO,
-                "[PF %p] could not create subchannel for address %s, ignoring",
-                policy_.get(), address.ToString().c_str());
-      }
+      GRPC_TRACE_LOG(pick_first, INFO)
+          << "[PF " << policy_.get()
+          << "] could not create subchannel for address " << address.ToString()
+          << ", ignoring";
       return;
     }
-    if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-      gpr_log(GPR_INFO,
-              "[PF %p] subchannel list %p index %" PRIuPTR
-              ": Created subchannel %p for address %s",
-              policy_.get(), this, subchannels_.size(), subchannel.get(),
-              address.ToString().c_str());
-    }
+    GRPC_TRACE_LOG(pick_first, INFO)
+        << "[PF " << policy_.get() << "] subchannel list " << this << " index "
+        << subchannels_.size() << ": Created subchannel " << subchannel.get()
+        << " for address " << address.ToString();
     subchannels_.emplace_back(std::make_unique<SubchannelData>(
         this, subchannels_.size(), std::move(subchannel)));
   });
 }
 
 PickFirst::SubchannelList::~SubchannelList() {
-  if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-    gpr_log(GPR_INFO, "[PF %p] Destroying subchannel_list %p", policy_.get(),
-            this);
-  }
+  GRPC_TRACE_LOG(pick_first, INFO)
+      << "[PF " << policy_.get() << "] Destroying subchannel_list " << this;
 }
 
 void PickFirst::SubchannelList::Orphan() {
-  if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-    gpr_log(GPR_INFO, "[PF %p] Shutting down subchannel_list %p", policy_.get(),
-            this);
-  }
+  GRPC_TRACE_LOG(pick_first, INFO)
+      << "[PF " << policy_.get() << "] Shutting down subchannel_list " << this;
   CHECK(!shutting_down_);
   shutting_down_ = true;
+  // Shut down subchannels.
+  subchannels_.clear();
   // Cancel Happy Eyeballs timer, if any.
   if (timer_handle_.has_value()) {
     policy_->channel_control_helper()->GetEventEngine()->Cancel(*timer_handle_);
@@ -1069,6 +1045,15 @@ void PickFirst::SubchannelList::ResetBackoffLocked() {
   }
 }
 
+void PickFirst::SubchannelList::ReportTransientFailure(absl::Status status) {
+  if (!resolution_note_.empty()) {
+    status = absl::Status(status.code(), absl::StrCat(status.message(), " (",
+                                                      resolution_note_, ")"));
+  }
+  policy_->UpdateState(GRPC_CHANNEL_TRANSIENT_FAILURE, status,
+                       MakeRefCounted<TransientFailurePicker>(status));
+}
+
 void PickFirst::SubchannelList::StartConnectingNextSubchannel() {
   // Find the next subchannel not in state TRANSIENT_FAILURE.
   // We skip subchannels in state TRANSIENT_FAILURE to avoid a
@@ -1082,6 +1067,7 @@ void PickFirst::SubchannelList::StartConnectingNextSubchannel() {
       sc->RequestConnectionWithTimer();
       return;
     }
+    sc->set_seen_transient_failure();
   }
   // If we didn't find a subchannel to request a connection on, check to
   // see if the Happy Eyeballs pass is complete.
@@ -1095,12 +1081,9 @@ void PickFirst::SubchannelList::MaybeFinishHappyEyeballsPass() {
   // We didn't find another subchannel not in state TRANSIENT_FAILURE,
   // so report TRANSIENT_FAILURE and switch to a mode in which we try to
   // connect to all addresses in parallel.
-  if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-    gpr_log(GPR_INFO,
-            "Pick First %p subchannel list %p failed to connect to "
-            "all subchannels",
-            policy_.get(), this);
-  }
+  GRPC_TRACE_LOG(pick_first, INFO)
+      << "Pick First " << policy_.get() << " subchannel list " << this
+      << " failed to connect to all subchannels";
   // Re-resolve and report TRANSIENT_FAILURE.
   policy_->channel_control_helper()->RequestReresolution();
   absl::Status status = absl::UnavailableError(
@@ -1108,8 +1091,7 @@ void PickFirst::SubchannelList::MaybeFinishHappyEyeballsPass() {
                         ? ""
                         : "failed to connect to all addresses; last error: "),
                    last_failure_.ToString()));
-  policy_->UpdateState(GRPC_CHANNEL_TRANSIENT_FAILURE, status,
-                       MakeRefCounted<TransientFailurePicker>(status));
+  ReportTransientFailure(std::move(status));
   // Drop the existing (working) connection, if any.  This may be
   // sub-optimal, but we can't ignore what the control plane told us.
   policy_->UnsetSelectedSubchannel();
@@ -1146,7 +1128,7 @@ class OldPickFirst final : public LoadBalancingPolicy {
                      RefCountedPtr<SubchannelInterface> subchannel);
 
       SubchannelInterface* subchannel() const { return subchannel_.get(); }
-      absl::optional<grpc_connectivity_state> connectivity_state() const {
+      std::optional<grpc_connectivity_state> connectivity_state() const {
         return connectivity_state_;
       }
       const absl::Status& connectivity_status() const {
@@ -1169,6 +1151,7 @@ class OldPickFirst final : public LoadBalancingPolicy {
       void ShutdownLocked();
 
       bool seen_transient_failure() const { return seen_transient_failure_; }
+      void set_seen_transient_failure() { seen_transient_failure_ = true; }
 
      private:
       // Watcher for subchannel connectivity state.
@@ -1216,14 +1199,14 @@ class OldPickFirst final : public LoadBalancingPolicy {
       SubchannelInterface::ConnectivityStateWatcherInterface* pending_watcher_ =
           nullptr;
       // Data updated by the watcher.
-      absl::optional<grpc_connectivity_state> connectivity_state_;
+      std::optional<grpc_connectivity_state> connectivity_state_;
       absl::Status connectivity_status_;
       bool seen_transient_failure_ = false;
     };
 
     SubchannelList(RefCountedPtr<OldPickFirst> policy,
                    EndpointAddressesIterator* addresses,
-                   const ChannelArgs& args);
+                   const ChannelArgs& args, absl::string_view resolution_note);
 
     ~SubchannelList() override;
 
@@ -1246,6 +1229,8 @@ class OldPickFirst final : public LoadBalancingPolicy {
       return true;
     }
 
+    void ReportTransientFailure(absl::Status status);
+
    private:
     // Returns true if all subchannels have seen their initial
     // connectivity state notifications.
@@ -1270,6 +1255,7 @@ class OldPickFirst final : public LoadBalancingPolicy {
     RefCountedPtr<OldPickFirst> policy_;
 
     ChannelArgs args_;
+    std::string resolution_note_;
 
     // The list of subchannels.
     std::vector<SubchannelData> subchannels_;
@@ -1286,7 +1272,7 @@ class OldPickFirst final : public LoadBalancingPolicy {
     // initial pass is over, this will be equal to size().
     size_t attempting_index_ = 0;
     // Happy Eyeballs timer handle.
-    absl::optional<grpc_event_engine::experimental::EventEngine::TaskHandle>
+    std::optional<grpc_event_engine::experimental::EventEngine::TaskHandle>
         timer_handle_;
 
     // After the initial Happy Eyeballs pass, the number of failures
@@ -1300,8 +1286,9 @@ class OldPickFirst final : public LoadBalancingPolicy {
   class HealthWatcher final
       : public SubchannelInterface::ConnectivityStateWatcherInterface {
    public:
-    explicit HealthWatcher(RefCountedPtr<OldPickFirst> policy)
-        : policy_(std::move(policy)) {}
+    HealthWatcher(RefCountedPtr<OldPickFirst> policy,
+                  absl::string_view resolution_note)
+        : policy_(std::move(policy)), resolution_note_(resolution_note) {}
 
     ~HealthWatcher() override {
       policy_.reset(DEBUG_LOCATION, "HealthWatcher dtor");
@@ -1316,6 +1303,7 @@ class OldPickFirst final : public LoadBalancingPolicy {
 
    private:
     RefCountedPtr<OldPickFirst> policy_;
+    std::string resolution_note_;
   };
 
   class Picker final : public SubchannelPicker {
@@ -1390,23 +1378,17 @@ OldPickFirst::OldPickFirst(Args args)
                     .GetInt(GRPC_ARG_HAPPY_EYEBALLS_CONNECTION_ATTEMPT_DELAY_MS)
                     .value_or(250),
                 100, 2000))) {
-  if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-    gpr_log(GPR_INFO, "Pick First %p created.", this);
-  }
+  GRPC_TRACE_LOG(pick_first, INFO) << "Pick First " << this << " created.";
 }
 
 OldPickFirst::~OldPickFirst() {
-  if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-    gpr_log(GPR_INFO, "Destroying Pick First %p", this);
-  }
+  GRPC_TRACE_LOG(pick_first, INFO) << "Destroying Pick First " << this;
   CHECK(subchannel_list_ == nullptr);
   CHECK(latest_pending_subchannel_list_ == nullptr);
 }
 
 void OldPickFirst::ShutdownLocked() {
-  if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-    gpr_log(GPR_INFO, "Pick First %p Shutting down", this);
-  }
+  GRPC_TRACE_LOG(pick_first, INFO) << "Pick First " << this << " Shutting down";
   shutdown_ = true;
   UnsetSelectedSubchannel();
   subchannel_list_.reset();
@@ -1416,9 +1398,8 @@ void OldPickFirst::ShutdownLocked() {
 void OldPickFirst::ExitIdleLocked() {
   if (shutdown_) return;
   if (IsIdle()) {
-    if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-      gpr_log(GPR_INFO, "Pick First %p exiting idle", this);
-    }
+    GRPC_TRACE_LOG(pick_first, INFO)
+        << "Pick First " << this << " exiting idle";
     AttemptToConnectUsingLatestUpdateArgsLocked();
   }
 }
@@ -1437,46 +1418,44 @@ void OldPickFirst::AttemptToConnectUsingLatestUpdateArgsLocked() {
     addresses = latest_update_args_.addresses->get();
   }
   // Replace latest_pending_subchannel_list_.
-  if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace) &&
+  if (GRPC_TRACE_FLAG_ENABLED(pick_first) &&
       latest_pending_subchannel_list_ != nullptr) {
-    gpr_log(GPR_INFO,
-            "[PF %p] Shutting down previous pending subchannel list %p", this,
-            latest_pending_subchannel_list_.get());
+    LOG(INFO) << "[PF " << this
+              << "] Shutting down previous pending subchannel list "
+              << latest_pending_subchannel_list_.get();
   }
   latest_pending_subchannel_list_ = MakeOrphanable<SubchannelList>(
-      RefAsSubclass<OldPickFirst>(), addresses, latest_update_args_.args);
+      RefAsSubclass<OldPickFirst>(), addresses, latest_update_args_.args,
+      latest_update_args_.resolution_note);
   // Empty update or no valid subchannels.  Put the channel in
   // TRANSIENT_FAILURE and request re-resolution.
   if (latest_pending_subchannel_list_->size() == 0) {
     channel_control_helper()->RequestReresolution();
-    absl::Status status =
-        latest_update_args_.addresses.ok()
-            ? absl::UnavailableError(absl::StrCat(
-                  "empty address list: ", latest_update_args_.resolution_note))
-            : latest_update_args_.addresses.status();
-    UpdateState(GRPC_CHANNEL_TRANSIENT_FAILURE, status,
-                MakeRefCounted<TransientFailurePicker>(status));
+    absl::Status status = latest_update_args_.addresses.ok()
+                              ? absl::UnavailableError("empty address list")
+                              : latest_update_args_.addresses.status();
+    latest_pending_subchannel_list_->ReportTransientFailure(std::move(status));
   }
   // If the new update is empty or we don't yet have a selected subchannel in
   // the current list, replace the current subchannel list immediately.
   if (latest_pending_subchannel_list_->size() == 0 || selected_ == nullptr) {
     UnsetSelectedSubchannel();
-    if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace) &&
-        subchannel_list_ != nullptr) {
-      gpr_log(GPR_INFO, "[PF %p] Shutting down previous subchannel list %p",
-              this, subchannel_list_.get());
+    if (GRPC_TRACE_FLAG_ENABLED(pick_first) && subchannel_list_ != nullptr) {
+      LOG(INFO) << "[PF " << this << "] Shutting down previous subchannel list "
+                << subchannel_list_.get();
     }
     subchannel_list_ = std::move(latest_pending_subchannel_list_);
   }
 }
 
 absl::Status OldPickFirst::UpdateLocked(UpdateArgs args) {
-  if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
+  if (GRPC_TRACE_FLAG_ENABLED(pick_first)) {
     if (args.addresses.ok()) {
-      gpr_log(GPR_INFO, "Pick First %p received update", this);
+      LOG(INFO) << "Pick First " << this << " received update";
     } else {
-      gpr_log(GPR_INFO, "Pick First %p received update with address error: %s",
-              this, args.addresses.status().ToString().c_str());
+      LOG(INFO) << "Pick First " << this
+                << " received update with address error: "
+                << args.addresses.status();
     }
   }
   // Set return status based on the address list.
@@ -1572,11 +1551,10 @@ void OldPickFirst::UnsetSelectedSubchannel() {
 void OldPickFirst::HealthWatcher::OnConnectivityStateChange(
     grpc_connectivity_state new_state, absl::Status status) {
   if (policy_->health_watcher_ != this) return;
-  if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-    gpr_log(GPR_INFO, "[PF %p] health watch state update: %s (%s)",
-            policy_.get(), ConnectivityStateName(new_state),
-            status.ToString().c_str());
-  }
+  GRPC_TRACE_LOG(pick_first, INFO)
+      << "[PF " << policy_.get()
+      << "] health watch state update: " << ConnectivityStateName(new_state)
+      << " (" << status << ")";
   switch (new_state) {
     case GRPC_CHANNEL_READY:
       policy_->channel_control_helper()->UpdateState(
@@ -1594,11 +1572,17 @@ void OldPickFirst::HealthWatcher::OnConnectivityStateChange(
           new_state, absl::OkStatus(),
           MakeRefCounted<QueuePicker>(policy_->Ref()));
       break;
-    case GRPC_CHANNEL_TRANSIENT_FAILURE:
+    case GRPC_CHANNEL_TRANSIENT_FAILURE: {
+      std::string message = absl::StrCat("health watch: ", status.message());
+      if (!resolution_note_.empty()) {
+        absl::StrAppend(&message, " (", resolution_note_, ")");
+      }
       policy_->channel_control_helper()->UpdateState(
           GRPC_CHANNEL_TRANSIENT_FAILURE, status,
-          MakeRefCounted<TransientFailurePicker>(status));
+          MakeRefCounted<TransientFailurePicker>(
+              absl::UnavailableError(message)));
       break;
+    }
     case GRPC_CHANNEL_SHUTDOWN:
       Crash("health watcher reported state SHUTDOWN");
   }
@@ -1614,13 +1598,10 @@ OldPickFirst::SubchannelList::SubchannelData::SubchannelData(
     : subchannel_list_(subchannel_list),
       index_(index),
       subchannel_(std::move(subchannel)) {
-  if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-    gpr_log(GPR_INFO,
-            "[PF %p] subchannel list %p index %" PRIuPTR
-            " (subchannel %p): starting watch",
-            subchannel_list_->policy_.get(), subchannel_list_, index_,
-            subchannel_.get());
-  }
+  GRPC_TRACE_LOG(pick_first, INFO)
+      << "[PF " << subchannel_list_->policy_.get() << "] subchannel list "
+      << subchannel_list_ << " index " << index_ << " (subchannel "
+      << subchannel_.get() << "): starting watch";
   auto watcher = std::make_unique<Watcher>(
       subchannel_list_->Ref(DEBUG_LOCATION, "Watcher"), index_);
   pending_watcher_ = watcher.get();
@@ -1629,13 +1610,11 @@ OldPickFirst::SubchannelList::SubchannelData::SubchannelData(
 
 void OldPickFirst::SubchannelList::SubchannelData::ShutdownLocked() {
   if (subchannel_ != nullptr) {
-    if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-      gpr_log(GPR_INFO,
-              "[PF %p] subchannel list %p index %" PRIuPTR " of %" PRIuPTR
-              " (subchannel %p): cancelling watch and unreffing subchannel",
-              subchannel_list_->policy_.get(), subchannel_list_, index_,
-              subchannel_list_->size(), subchannel_.get());
-    }
+    GRPC_TRACE_LOG(pick_first, INFO)
+        << "[PF " << subchannel_list_->policy_.get() << "] subchannel list "
+        << subchannel_list_ << " index " << index_ << " of "
+        << subchannel_list_->size() << " (subchannel " << subchannel_.get()
+        << "): cancelling watch and unreffing subchannel";
     subchannel_->CancelConnectivityStateWatch(pending_watcher_);
     pending_watcher_ = nullptr;
     subchannel_.reset();
@@ -1645,24 +1624,22 @@ void OldPickFirst::SubchannelList::SubchannelData::ShutdownLocked() {
 void OldPickFirst::SubchannelList::SubchannelData::OnConnectivityStateChange(
     grpc_connectivity_state new_state, absl::Status status) {
   OldPickFirst* p = subchannel_list_->policy_.get();
-  if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-    gpr_log(
-        GPR_INFO,
-        "[PF %p] subchannel list %p index %" PRIuPTR " of %" PRIuPTR
-        " (subchannel %p): connectivity changed: old_state=%s, new_state=%s, "
-        "status=%s, shutting_down=%d, pending_watcher=%p, "
-        "seen_transient_failure=%d, p->selected_=%p, "
-        "p->subchannel_list_=%p, p->latest_pending_subchannel_list_=%p",
-        p, subchannel_list_, index_, subchannel_list_->size(),
-        subchannel_.get(),
-        (connectivity_state_.has_value()
-             ? ConnectivityStateName(*connectivity_state_)
-             : "N/A"),
-        ConnectivityStateName(new_state), status.ToString().c_str(),
-        subchannel_list_->shutting_down_, pending_watcher_,
-        seen_transient_failure_, p->selected_, p->subchannel_list_.get(),
-        p->latest_pending_subchannel_list_.get());
-  }
+  GRPC_TRACE_LOG(pick_first, INFO)
+      << "[PF " << p << "] subchannel list " << subchannel_list_ << " index "
+      << index_ << " of " << subchannel_list_->size() << " (subchannel "
+      << subchannel_.get() << "): connectivity changed: old_state="
+      << (connectivity_state_.has_value()
+              ? ConnectivityStateName(*connectivity_state_)
+              : "N/A")
+      << ", new_state=" << ConnectivityStateName(new_state)
+      << ", status=" << status
+      << ", shutting_down=" << subchannel_list_->shutting_down_
+      << ", pending_watcher=" << pending_watcher_
+      << ", seen_transient_failure=" << seen_transient_failure_
+      << ", p->selected_=" << p->selected_
+      << ", p->subchannel_list_=" << p->subchannel_list_.get()
+      << ", p->latest_pending_subchannel_list_="
+      << p->latest_pending_subchannel_list_.get();
   if (subchannel_list_->shutting_down_ || pending_watcher_ == nullptr) return;
   auto& stats_plugins = subchannel_list_->policy_->channel_control_helper()
                             ->GetStatsPluginGroup();
@@ -1671,17 +1648,15 @@ void OldPickFirst::SubchannelList::SubchannelData::OnConnectivityStateChange(
   CHECK(subchannel_list_ == p->subchannel_list_.get() ||
         subchannel_list_ == p->latest_pending_subchannel_list_.get());
   CHECK(new_state != GRPC_CHANNEL_SHUTDOWN);
-  absl::optional<grpc_connectivity_state> old_state = connectivity_state_;
+  std::optional<grpc_connectivity_state> old_state = connectivity_state_;
   connectivity_state_ = new_state;
   connectivity_status_ = std::move(status);
   // Handle updates for the currently selected subchannel.
   if (p->selected_ == this) {
     CHECK(subchannel_list_ == p->subchannel_list_.get());
-    if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-      gpr_log(GPR_INFO,
-              "Pick First %p selected subchannel connectivity changed to %s", p,
-              ConnectivityStateName(new_state));
-    }
+    GRPC_TRACE_LOG(pick_first, INFO)
+        << "Pick First " << p << " selected subchannel connectivity changed to "
+        << ConnectivityStateName(new_state);
     // Any state change is considered to be a failure of the existing
     // connection.
     stats_plugins.AddCounter(
@@ -1696,13 +1671,10 @@ void OldPickFirst::SubchannelList::SubchannelData::OnConnectivityStateChange(
     p->channel_control_helper()->RequestReresolution();
     // If there is a pending update, switch to the pending update.
     if (p->latest_pending_subchannel_list_ != nullptr) {
-      if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-        gpr_log(GPR_INFO,
-                "Pick First %p promoting pending subchannel list %p to "
-                "replace %p",
-                p, p->latest_pending_subchannel_list_.get(),
-                p->subchannel_list_.get());
-      }
+      GRPC_TRACE_LOG(pick_first, INFO)
+          << "Pick First " << p << " promoting pending subchannel list "
+          << p->latest_pending_subchannel_list_.get() << " to replace "
+          << p->subchannel_list_.get();
       p->UnsetSelectedSubchannel();
       p->subchannel_list_ = std::move(p->latest_pending_subchannel_list_);
       // Set our state to that of the pending subchannel list.
@@ -1711,10 +1683,9 @@ void OldPickFirst::SubchannelList::SubchannelData::OnConnectivityStateChange(
             "selected subchannel failed; switching to pending update; "
             "last failure: ",
             p->subchannel_list_->last_failure_.ToString()));
-        p->UpdateState(GRPC_CHANNEL_TRANSIENT_FAILURE, status,
-                       MakeRefCounted<TransientFailurePicker>(status));
+        subchannel_list_->ReportTransientFailure(std::move(status));
       } else if (p->state_ != GRPC_CHANNEL_TRANSIENT_FAILURE) {
-        p->UpdateState(GRPC_CHANNEL_CONNECTING, absl::Status(),
+        p->UpdateState(GRPC_CHANNEL_CONNECTING, absl::OkStatus(),
                        MakeRefCounted<QueuePicker>(nullptr));
       }
       return;
@@ -1723,7 +1694,7 @@ void OldPickFirst::SubchannelList::SubchannelData::OnConnectivityStateChange(
     p->UnsetSelectedSubchannel();
     p->subchannel_list_.reset();
     p->UpdateState(
-        GRPC_CHANNEL_IDLE, absl::Status(),
+        GRPC_CHANNEL_IDLE, absl::OkStatus(),
         MakeRefCounted<QueuePicker>(p->Ref(DEBUG_LOCATION, "QueuePicker")));
     return;
   }
@@ -1751,10 +1722,8 @@ void OldPickFirst::SubchannelList::SubchannelData::OnConnectivityStateChange(
     ProcessUnselectedReadyLocked();
     return;
   }
-  // Make sure we note when a subchannel has seen TRANSIENT_FAILURE.
-  bool prev_seen_transient_failure = seen_transient_failure_;
+  // Record status for TRANSIENT_FAILURE state.
   if (new_state == GRPC_CHANNEL_TRANSIENT_FAILURE) {
-    seen_transient_failure_ = true;
     subchannel_list_->last_failure_ = connectivity_status_;
   }
   // If this is the initial connectivity state update for this subchannel,
@@ -1783,6 +1752,8 @@ void OldPickFirst::SubchannelList::SubchannelData::OnConnectivityStateChange(
   // Otherwise, process connectivity state change.
   switch (*connectivity_state_) {
     case GRPC_CHANNEL_TRANSIENT_FAILURE: {
+      bool prev_seen_transient_failure =
+          std::exchange(seen_transient_failure_, true);
       // If this is the first failure we've seen on this subchannel,
       // then we're still in the Happy Eyeballs pass.
       if (!prev_seen_transient_failure && seen_transient_failure_) {
@@ -1820,8 +1791,7 @@ void OldPickFirst::SubchannelList::SubchannelData::OnConnectivityStateChange(
                    ? ""
                    : "failed to connect to all addresses; last error: "),
               connectivity_status_.ToString()));
-          p->UpdateState(GRPC_CHANNEL_TRANSIENT_FAILURE, status,
-                         MakeRefCounted<TransientFailurePicker>(status));
+          subchannel_list_->ReportTransientFailure(std::move(status));
         }
       }
       break;
@@ -1839,7 +1809,7 @@ void OldPickFirst::SubchannelList::SubchannelData::OnConnectivityStateChange(
       // already in TRANSIENT_FAILURE.
       if (subchannel_list_ == p->subchannel_list_.get() &&
           p->state_ != GRPC_CHANNEL_TRANSIENT_FAILURE) {
-        p->UpdateState(GRPC_CHANNEL_CONNECTING, absl::Status(),
+        p->UpdateState(GRPC_CHANNEL_CONNECTING, absl::OkStatus(),
                        MakeRefCounted<QueuePicker>(nullptr));
       }
       break;
@@ -1860,39 +1830,31 @@ void OldPickFirst::SubchannelList::SubchannelData::
   // If this is not the last subchannel in the list, start the timer.
   if (index_ != subchannel_list_->size() - 1) {
     OldPickFirst* p = subchannel_list_->policy_.get();
-    if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-      gpr_log(GPR_INFO,
-              "Pick First %p subchannel list %p: starting Connection "
-              "Attempt Delay timer for %" PRId64 "ms for index %" PRIuPTR,
-              p, subchannel_list_, p->connection_attempt_delay_.millis(),
-              index_);
-    }
+    GRPC_TRACE_LOG(pick_first, INFO)
+        << "Pick First " << p << " subchannel list " << subchannel_list_
+        << ": starting Connection Attempt Delay timer for "
+        << p->connection_attempt_delay_.millis() << "ms for index " << index_;
     subchannel_list_->timer_handle_ =
         p->channel_control_helper()->GetEventEngine()->RunAfter(
             p->connection_attempt_delay_,
             [subchannel_list =
                  subchannel_list_->Ref(DEBUG_LOCATION, "timer")]() mutable {
-              ApplicationCallbackExecCtx application_exec_ctx;
               ExecCtx exec_ctx;
               auto* sl = subchannel_list.get();
               sl->policy_->work_serializer()->Run(
                   [subchannel_list = std::move(subchannel_list)]() {
-                    if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-                      gpr_log(GPR_INFO,
-                              "Pick First %p subchannel list %p: Connection "
-                              "Attempt Delay timer fired (shutting_down=%d, "
-                              "selected=%p)",
-                              subchannel_list->policy_.get(),
-                              subchannel_list.get(),
-                              subchannel_list->shutting_down_,
-                              subchannel_list->policy_->selected_);
-                    }
+                    GRPC_TRACE_LOG(pick_first, INFO)
+                        << "Pick First " << subchannel_list->policy_.get()
+                        << " subchannel list " << subchannel_list.get()
+                        << ": Connection Attempt Delay timer fired "
+                        << "(shutting_down=" << subchannel_list->shutting_down_
+                        << ", selected=" << subchannel_list->policy_->selected_
+                        << ")";
                     if (subchannel_list->shutting_down_) return;
                     if (subchannel_list->policy_->selected_ != nullptr) return;
                     ++subchannel_list->attempting_index_;
                     subchannel_list->StartConnectingNextSubchannel();
-                  },
-                  DEBUG_LOCATION);
+                  });
             });
   }
 }
@@ -1918,39 +1880,34 @@ void OldPickFirst::SubchannelList::SubchannelData::
         subchannel_list_ == p->latest_pending_subchannel_list_.get());
   // Case 2.  Promote p->latest_pending_subchannel_list_ to p->subchannel_list_.
   if (subchannel_list_ == p->latest_pending_subchannel_list_.get()) {
-    if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-      gpr_log(GPR_INFO,
-              "Pick First %p promoting pending subchannel list %p to "
-              "replace %p",
-              p, p->latest_pending_subchannel_list_.get(),
-              p->subchannel_list_.get());
-    }
+    GRPC_TRACE_LOG(pick_first, INFO)
+        << "Pick First " << p << " promoting pending subchannel list "
+        << p->latest_pending_subchannel_list_.get() << " to replace "
+        << p->subchannel_list_.get();
     p->UnsetSelectedSubchannel();
     p->subchannel_list_ = std::move(p->latest_pending_subchannel_list_);
   }
   // Cases 1 and 2.
-  if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-    gpr_log(GPR_INFO, "Pick First %p selected subchannel %p", p,
-            subchannel_.get());
-  }
+  GRPC_TRACE_LOG(pick_first, INFO)
+      << "Pick First " << p << " selected subchannel " << subchannel_.get();
   p->selected_ = this;
   // If health checking is enabled, start the health watch, but don't
   // report a new picker -- we want to stay in CONNECTING while we wait
   // for the health status notification.
   // If health checking is NOT enabled, report READY.
   if (p->enable_health_watch_) {
-    if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-      gpr_log(GPR_INFO, "[PF %p] starting health watch", p);
-    }
+    GRPC_TRACE_LOG(pick_first, INFO)
+        << "[PF " << p << "] starting health watch";
     auto watcher = std::make_unique<HealthWatcher>(
-        p->RefAsSubclass<OldPickFirst>(DEBUG_LOCATION, "HealthWatcher"));
+        p->RefAsSubclass<OldPickFirst>(DEBUG_LOCATION, "HealthWatcher"),
+        subchannel_list_->resolution_note_);
     p->health_watcher_ = watcher.get();
     auto health_data_watcher = MakeHealthCheckWatcher(
         p->work_serializer(), subchannel_list_->args_, std::move(watcher));
     p->health_data_watcher_ = health_data_watcher.get();
     subchannel_->AddDataWatcher(std::move(health_data_watcher));
   } else {
-    p->UpdateState(GRPC_CHANNEL_READY, absl::Status(),
+    p->UpdateState(GRPC_CHANNEL_READY, absl::OkStatus(),
                    MakeRefCounted<Picker>(subchannel()->Ref()));
   }
   // Unref all other subchannels in the list.
@@ -1967,18 +1924,17 @@ void OldPickFirst::SubchannelList::SubchannelData::
 
 OldPickFirst::SubchannelList::SubchannelList(
     RefCountedPtr<OldPickFirst> policy, EndpointAddressesIterator* addresses,
-    const ChannelArgs& args)
+    const ChannelArgs& args, absl::string_view resolution_note)
     : InternallyRefCounted<SubchannelList>(
-          GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace) ? "SubchannelList"
-                                                            : nullptr),
+          GRPC_TRACE_FLAG_ENABLED(pick_first) ? "SubchannelList" : nullptr),
       policy_(std::move(policy)),
-      args_(args.Remove(GRPC_ARG_INTERNAL_PICK_FIRST_ENABLE_HEALTH_CHECKING)
-                .Remove(
-                    GRPC_ARG_INTERNAL_PICK_FIRST_OMIT_STATUS_MESSAGE_PREFIX)) {
-  if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-    gpr_log(GPR_INFO, "[PF %p] Creating subchannel list %p - channel args: %s",
-            policy_.get(), this, args_.ToString().c_str());
-  }
+      args_(
+          args.Remove(GRPC_ARG_INTERNAL_PICK_FIRST_ENABLE_HEALTH_CHECKING)
+              .Remove(GRPC_ARG_INTERNAL_PICK_FIRST_OMIT_STATUS_MESSAGE_PREFIX)),
+      resolution_note_(resolution_note) {
+  GRPC_TRACE_LOG(pick_first, INFO)
+      << "[PF " << policy_.get() << "] Creating subchannel list " << this
+      << " - channel args: " << args_.ToString();
   if (addresses == nullptr) return;
   // Create a subchannel for each address.
   addresses->ForEach([&](const EndpointAddresses& address) {
@@ -1988,36 +1944,28 @@ OldPickFirst::SubchannelList::SubchannelList(
             address.address(), address.args(), args_);
     if (subchannel == nullptr) {
       // Subchannel could not be created.
-      if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-        gpr_log(GPR_INFO,
-                "[PF %p] could not create subchannel for address %s, ignoring",
-                policy_.get(), address.ToString().c_str());
-      }
+      GRPC_TRACE_LOG(pick_first, INFO)
+          << "[PF " << policy_.get()
+          << "] could not create subchannel for address " << address.ToString()
+          << ", ignoring";
       return;
     }
-    if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-      gpr_log(GPR_INFO,
-              "[PF %p] subchannel list %p index %" PRIuPTR
-              ": Created subchannel %p for address %s",
-              policy_.get(), this, subchannels_.size(), subchannel.get(),
-              address.ToString().c_str());
-    }
+    GRPC_TRACE_LOG(pick_first, INFO)
+        << "[PF " << policy_.get() << "] subchannel list " << this << " index "
+        << subchannels_.size() << ": Created subchannel " << subchannel.get()
+        << " for address " << address.ToString();
     subchannels_.emplace_back(this, subchannels_.size(), std::move(subchannel));
   });
 }
 
 OldPickFirst::SubchannelList::~SubchannelList() {
-  if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-    gpr_log(GPR_INFO, "[PF %p] Destroying subchannel_list %p", policy_.get(),
-            this);
-  }
+  GRPC_TRACE_LOG(pick_first, INFO)
+      << "[PF " << policy_.get() << "] Destroying subchannel_list " << this;
 }
 
 void OldPickFirst::SubchannelList::Orphan() {
-  if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-    gpr_log(GPR_INFO, "[PF %p] Shutting down subchannel_list %p", policy_.get(),
-            this);
-  }
+  GRPC_TRACE_LOG(pick_first, INFO)
+      << "[PF " << policy_.get() << "] Shutting down subchannel_list " << this;
   CHECK(!shutting_down_);
   shutting_down_ = true;
   for (auto& sd : subchannels_) {
@@ -2035,6 +1983,15 @@ void OldPickFirst::SubchannelList::ResetBackoffLocked() {
   }
 }
 
+void OldPickFirst::SubchannelList::ReportTransientFailure(absl::Status status) {
+  if (!resolution_note_.empty()) {
+    status = absl::Status(status.code(), absl::StrCat(status.message(), " (",
+                                                      resolution_note_, ")"));
+  }
+  policy_->UpdateState(GRPC_CHANNEL_TRANSIENT_FAILURE, status,
+                       MakeRefCounted<TransientFailurePicker>(status));
+}
+
 void OldPickFirst::SubchannelList::StartConnectingNextSubchannel() {
   // Find the next subchannel not in state TRANSIENT_FAILURE.
   // We skip subchannels in state TRANSIENT_FAILURE to avoid a
@@ -2048,6 +2005,7 @@ void OldPickFirst::SubchannelList::StartConnectingNextSubchannel() {
       sc->RequestConnectionWithTimer();
       return;
     }
+    sc->set_seen_transient_failure();
   }
   // If we didn't find a subchannel to request a connection on, check to
   // see if the Happy Eyeballs pass is complete.
@@ -2061,23 +2019,18 @@ void OldPickFirst::SubchannelList::MaybeFinishHappyEyeballsPass() {
   // We didn't find another subchannel not in state TRANSIENT_FAILURE,
   // so report TRANSIENT_FAILURE and switch to a mode in which we try to
   // connect to all addresses in parallel.
-  if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-    gpr_log(GPR_INFO,
-            "Pick First %p subchannel list %p failed to connect to "
-            "all subchannels",
-            policy_.get(), this);
-  }
+  GRPC_TRACE_LOG(pick_first, INFO)
+      << "Pick First " << policy_.get() << " subchannel list " << this
+      << " failed to connect to all subchannels";
   // In case 2, swap to the new subchannel list.  This means reporting
   // TRANSIENT_FAILURE and dropping the existing (working) connection,
   // but we can't ignore what the control plane has told us.
   if (policy_->latest_pending_subchannel_list_.get() == this) {
-    if (GRPC_TRACE_FLAG_ENABLED(grpc_lb_pick_first_trace)) {
-      gpr_log(GPR_INFO,
-              "Pick First %p promoting pending subchannel list %p to "
-              "replace %p",
-              policy_.get(), policy_->latest_pending_subchannel_list_.get(),
-              this);
-    }
+    GRPC_TRACE_LOG(pick_first, INFO)
+        << "Pick First " << policy_.get()
+        << " promoting pending subchannel list "
+        << policy_->latest_pending_subchannel_list_.get() << " to replace "
+        << this;
     policy_->UnsetSelectedSubchannel();
     policy_->subchannel_list_ =
         std::move(policy_->latest_pending_subchannel_list_);
@@ -2092,8 +2045,7 @@ void OldPickFirst::SubchannelList::MaybeFinishHappyEyeballsPass() {
                           ? ""
                           : "failed to connect to all addresses; last error: "),
                      last_failure_.ToString()));
-    policy_->UpdateState(GRPC_CHANNEL_TRANSIENT_FAILURE, status,
-                         MakeRefCounted<TransientFailurePicker>(status));
+    ReportTransientFailure(std::move(status));
   }
   // We now transition into a mode where we try to connect to all
   // subchannels in parallel.  For any subchannel currently in IDLE,