diff --git a/include/app_metrics.h b/include/app_metrics.h
index a84fe3fc..47cf5546 100644
--- a/include/app_metrics.h
+++ b/include/app_metrics.h
@@ -50,6 +50,7 @@ public:
     static inline const std::string DOC_WRITE_LABEL = "write";
     static inline const std::string IMPORT_LABEL = "import";
     static inline const std::string DOC_DELETE_LABEL = "delete";
+    static inline const std::string OVERLOADED_LABEL = "overloaded";
 
     static const uint64_t METRICS_REFRESH_INTERVAL_MS = 10 * 1000;
 
diff --git a/include/http_data.h b/include/http_data.h
index 4c5dc14c..71a0e819 100644
--- a/include/http_data.h
+++ b/include/http_data.h
@@ -235,6 +235,9 @@ struct http_req {
     // timestamp from the underlying http library
     uint64_t conn_ts;
 
+    // was the request aborted *without a result* because of wait time exceeding search cutoff threshold?
+    bool overloaded = false;
+
     std::mutex mcv;
     std::condition_variable cv;
     bool ready;
@@ -301,7 +304,9 @@ struct http_req {
             bool log_slow_requests = config.get_log_slow_requests_time_ms() >= 0 &&
                                      int(ms_since_start) >= config.get_log_slow_requests_time_ms();
 
-            if(log_slow_searches || log_slow_requests) {
+            if(overloaded) {
+                AppMetrics::get_instance().increment_count(AppMetrics::OVERLOADED_LABEL, 1);
+            } else if(log_slow_searches || log_slow_requests) {
                 // log slow request if logging is enabled
                 std::string query_string = "?";
                 bool is_multi_search_query = (path_without_query == "/multi_search");
diff --git a/src/app_metrics.cpp b/src/app_metrics.cpp
index b4c6c299..46523b2f 100644
--- a/src/app_metrics.cpp
+++ b/src/app_metrics.cpp
@@ -35,6 +35,8 @@ void AppMetrics::get(const std::string& rps_key, const std::string& latency_key,
     auto DOC_DELETE_RPS_KEY = DOC_DELETE_LABEL + "_" + rps_key;
     auto DOC_DELETE_LATENCY_KEY = DOC_DELETE_LABEL + "_" + latency_key;
 
+    auto OVERLOADED_RPS_KEY = OVERLOADED_LABEL + "_" + rps_key;
+
     result[rps_key] = nlohmann::json::object();
     for(const auto& kv: *counts) {
         if(kv.first == SEARCH_LABEL) {
@@ -53,6 +55,10 @@ void AppMetrics::get(const std::string& rps_key, const std::string& latency_key,
             result[DOC_DELETE_RPS_KEY] = double(kv.second) / (METRICS_REFRESH_INTERVAL_MS / 1000);
         }
 
+        else if(kv.first == OVERLOADED_LABEL) {
+            result[OVERLOADED_RPS_KEY] = double(kv.second) / (METRICS_REFRESH_INTERVAL_MS / 1000);
+        }
+
         else {
             result[rps_key][kv.first] = (double(kv.second) / (METRICS_REFRESH_INTERVAL_MS / 1000));
             total_counts += kv.second;
@@ -90,7 +96,8 @@ void AppMetrics::get(const std::string& rps_key, const std::string& latency_key,
 
     std::vector<std::string> keys_to_check = {
         SEARCH_RPS_KEY, IMPORT_RPS_KEY, DOC_WRITE_RPS_KEY, DOC_DELETE_RPS_KEY,
-        SEARCH_LATENCY_KEY, IMPORT_LATENCY_KEY, DOC_WRITE_LATENCY_KEY, DOC_DELETE_LATENCY_KEY
+        SEARCH_LATENCY_KEY, IMPORT_LATENCY_KEY, DOC_WRITE_LATENCY_KEY, DOC_DELETE_LATENCY_KEY,
+        OVERLOADED_RPS_KEY
     };
 
     for(auto& key: keys_to_check) {
diff --git a/src/collection.cpp b/src/collection.cpp
index f514242a..1b05ff12 100644
--- a/src/collection.cpp
+++ b/src/collection.cpp
@@ -1299,6 +1299,12 @@ Option<nlohmann::json> Collection::search(const std::string & raw_query,
         total_found = search_params->all_result_ids_len;
     }
 
+    if(search_cutoff && total_found == 0) {
+        // this can happen if other requests stopped this request from being processed
+        // we should return an error so that request can be retried by client
+        return Option<nlohmann::json>(529, "Site is overloaded");
+    }
+
     if(match_score_index >= 0 && sort_fields_std[match_score_index].text_match_buckets > 1) {
         size_t num_buckets = sort_fields_std[match_score_index].text_match_buckets;
         const size_t max_kvs_bucketed = std::min<size_t>(DEFAULT_TOPSTER_SIZE, raw_result_kvs.size());
diff --git a/src/core_api.cpp b/src/core_api.cpp
index 07ec2022..31552463 100644
--- a/src/core_api.cpp
+++ b/src/core_api.cpp
@@ -392,6 +392,9 @@ bool get_search(const std::shared_ptr<http_req>& req, const std::shared_ptr<http
 
     if(!search_op.ok()) {
         res->set(search_op.code(), search_op.error());
+        if(search_op.code() == 529) {
+            req->overloaded = true;
+        }
         return false;
     }
 
@@ -562,6 +565,11 @@ bool post_multi_search(const std::shared_ptr<http_req>& req, const std::shared_p
         if(search_op.ok()) {
             response["results"].push_back(nlohmann::json::parse(results_json_str));
         } else {
+            if(search_op.code() == 529) {
+                res->set(search_op.code(), search_op.error());
+                req->overloaded = true;
+                return false;
+            }
             nlohmann::json err_res;
             err_res["error"] = search_op.error();
             err_res["code"] = search_op.code();