[service] improve per-instance metrics failure diagnostics

Include trace ID and instance ID when a per-instance service metrics collector cannot be resolved. Document the two possible failure scopes in the log so unknown-instance requests can be distinguished from the API business path without exposing request payloads.

Author: charpty

kv_cache_manager/service/util/common.h

@@ -34,7 +34,11 @@ std::string ExtractIpFromPeer(const std::string &peer);
 #define API_CONTEXT_GET_AND_INIT_COLLECTOR(method, return_value)                                                       \
     auto metrics_collector = get_metrics_collector_from_map_for_##method(request->instance_id());                      \
     if (metrics_collector == nullptr) {                                                                                \
-        KVCM_LOG_ERROR("get " #method " metrics collector failed");                                                    \
+        KVCM_LOG_ERROR("[traceId: %s] get " #method                                                                    \
+                       " metrics collector failed, instance_id: %s; instance is not registered or metrics collector "  \
+                       "initialization failed",                                                                        \
+                       request->trace_id().c_str(),                                                                    \
+                       request->instance_id().c_str());                                                                \
         auto *header = response->mutable_header();                                                                     \
         auto *status = header->mutable_status();                                                                       \
         status->set_code(proto::meta::INSTANCE_NOT_EXIST);                                                             \