diff --git a/Framework/Core/include/Framework/DataProcessingDevice.h b/Framework/Core/include/Framework/DataProcessingDevice.h
index 89d8dd90d9f20..0d00b99e6a03a 100644
--- a/Framework/Core/include/Framework/DataProcessingDevice.h
+++ b/Framework/Core/include/Framework/DataProcessingDevice.h
@@ -74,7 +74,6 @@ struct DataProcessorContext {
   AlgorithmSpec::ErrorCallback* error = nullptr;
 
   std::function<void(o2::framework::RuntimeErrorRef e, InputRecord& record)>* errorHandling = nullptr;
-  int* errorCount = nullptr;
 };
 
 /// A device actually carrying out all the DPL
@@ -124,7 +123,6 @@ class DataProcessingDevice : public FairMQDevice
   /// Completed actions
   std::vector<DataRelayer::RecordAction> mCompleted;
 
-  int mErrorCount;
   uint64_t mLastSlowMetricSentTimestamp = 0;         /// The timestamp of the last time we sent slow metrics
   uint64_t mLastMetricFlushedTimestamp = 0;          /// The timestamp of the last time we actually flushed metrics
   uint64_t mBeginIterationTimestamp = 0;             /// The timestamp of when the current ConditionalRun was started
diff --git a/Framework/Core/include/Framework/DataProcessingStats.h b/Framework/Core/include/Framework/DataProcessingStats.h
index b4d9fb5c26210..c5bde7ed5598e 100644
--- a/Framework/Core/include/Framework/DataProcessingStats.h
+++ b/Framework/Core/include/Framework/DataProcessingStats.h
@@ -25,6 +25,7 @@ struct DataProcessingStats {
     int minLatency = 0;
     int maxLatency = 0;
   };
+  std::atomic<int> errorCount = 0;
   std::atomic<int> pendingInputs = 0;
   std::atomic<int> incomplete = 0;
   std::atomic<int> inputParts = 0;
diff --git a/Framework/Core/src/CommonServices.cxx b/Framework/Core/src/CommonServices.cxx
index fa56c952c2509..40d074f894701 100644
--- a/Framework/Core/src/CommonServices.cxx
+++ b/Framework/Core/src/CommonServices.cxx
@@ -480,6 +480,7 @@ auto sendRelayerMetrics(ServiceRegistry& registry, DataProcessingStats& stats) -
   monitoring.send(Metric{(int)relayerStats.droppedIncomingMessages, "dropped_incoming_messages"}.addTag(Key::Subsystem, Value::DPL));
   monitoring.send(Metric{(int)relayerStats.relayedMessages, "relayed_messages"}.addTag(Key::Subsystem, Value::DPL));
 
+  monitoring.send(Metric{(int)stats.errorCount, "errors"}.addTag(Key::Subsystem, Value::DPL));
   monitoring.send(Metric{(int)stats.pendingInputs, "inputs/relayed/pending"}.addTag(Key::Subsystem, Value::DPL));
   monitoring.send(Metric{(int)stats.incomplete, "inputs/relayed/incomplete"}.addTag(Key::Subsystem, Value::DPL));
   monitoring.send(Metric{(int)stats.inputParts, "inputs/relayed/total"}.addTag(Key::Subsystem, Value::DPL));
diff --git a/Framework/Core/src/DataProcessingDevice.cxx b/Framework/Core/src/DataProcessingDevice.cxx
index da5982055957d..139f4d309eb69 100644
--- a/Framework/Core/src/DataProcessingDevice.cxx
+++ b/Framework/Core/src/DataProcessingDevice.cxx
@@ -94,8 +94,7 @@ DataProcessingDevice::DataProcessingDevice(DeviceSpec const& spec, ServiceRegist
     mError{spec.algorithm.onError},
     mConfigRegistry{nullptr},
     mAllocator{&mTimingInfo, &registry, spec.outputs},
-    mServiceRegistry{registry},
-    mErrorCount{0}
+    mServiceRegistry{registry}
 {
   /// FIXME: move erro handling to a service?
   if (mError != nullptr) {
@@ -427,7 +426,6 @@ void DataProcessingDevice::fillContext(DataProcessorContext& context)
   context.error = &mError;
   /// Callback for the error handling
   context.errorHandling = &mErrorHandling;
-  context.errorCount = &mErrorCount;
 }
 
 void DataProcessingDevice::PreRun()
@@ -684,9 +682,8 @@ void DataProcessingDevice::handleData(DataProcessorContext& context, FairMQParts
     return results;
   };
 
-  auto reportError = [& registry = *context.registry, &context](const char* message) {
-    context.errorCount++;
-    registry.get<Monitoring>().send(Metric{*context.errorCount, "errors"}.addTag(Key::Subsystem, Value::DPL));
+  auto reportError = [&registry = *context.registry, &context](const char* message) {
+    registry.get<DataProcessingStats>().errorCount++;
   };
 
   auto handleValidMessages = [&parts, &context = context, &relayer = *context.relayer, &reportError](std::vector<InputType> const& types) {
@@ -782,9 +779,8 @@ bool DataProcessingDevice::tryDispatchComputation(DataProcessorContext& context,
   // should work just fine.
   std::vector<MessageSet> currentSetOfInputs;
 
-  auto reportError = [& registry = *context.registry, &context](const char* message) {
-    context.errorCount++;
-    registry.get<Monitoring>().send(Metric{*context.errorCount, "errors"}.addTag(Key::Subsystem, Value::DPL));
+  auto reportError = [&registry = *context.registry, &context](const char* message) {
+    registry.get<DataProcessingStats>().errorCount++;
   };
 
   // For the moment we have a simple "immediately dispatch" policy for stuff
@@ -1077,8 +1073,7 @@ bool DataProcessingDevice::tryDispatchComputation(DataProcessorContext& context,
 void DataProcessingDevice::error(const char* msg)
 {
   LOG(ERROR) << msg;
-  mErrorCount++;
-  mServiceRegistry.get<Monitoring>().send(Metric{mErrorCount, "errors"}.addTag(Key::Subsystem, Value::DPL));
+  mServiceRegistry.get<DataProcessingStats>().errorCount++;
 }
 
 } // namespace o2::framework