Simple server wide memory profiler

programs/server/Server.cpp

@@ -617,9 +617,19 @@ int Server::main(const std::vector<std::string> & /*args*/)
     /// Look at compiler-rt/lib/sanitizer_common/sanitizer_stacktrace.h
     ///
 #if USE_UNWIND && !WITH_COVERAGE && !defined(SANITIZER)
-    /// QueryProfiler cannot work reliably with any other libunwind or without PHDR cache.
+    /// Profilers cannot work reliably with any other libunwind or without PHDR cache.
     if (hasPHDRCache())
+    {
         global_context->initializeTraceCollector();
+
+        /// Set up server-wide memory profiler (for total memory tracker).
+        UInt64 total_memory_profiler_step = config().getUInt64("total_memory_profiler_step", 0);
+        if (total_memory_profiler_step)
+        {
+            total_memory_tracker.setOrRaiseProfilerLimit(total_memory_profiler_step);
+            total_memory_tracker.setProfilerStep(total_memory_profiler_step);
+        }
+    }
 #endif
 
     /// Describe multiple reasons when query profiler cannot work.

programs/server/config.xml

@@ -107,6 +107,12 @@
       -->
     <max_server_memory_usage_to_ram_ratio>0.9</max_server_memory_usage_to_ram_ratio>
 
+    <!-- Simple server-wide memory profiler. Collect a stack trace at every peak allocation step (in bytes).
+         Data will be stored in system.trace_log table with query_id = empty string.
+         Zero means disabled. Minimal effective value is 4 MiB.
+      -->
+    <total_memory_profiler_step>4194304</total_memory_profiler_step>
+
     <!-- Set limit on number of open files (default: maximum). This setting makes sense on Mac OS X because getrlimit() fails to retrieve
          correct maximum value. -->
     <!-- <max_open_files>262144</max_open_files> -->

src/Common/TraceCollector.cpp

@@ -65,7 +65,7 @@ void TraceCollector::collect(TraceType trace_type, const StackTrace & stack_trac
         sizeof(StackTrace::Frames) +           // collected stack trace, maximum capacity
         sizeof(TraceType) +                    // trace type
         sizeof(UInt64) +                       // thread_id
-        sizeof(UInt64);                         // size
+        sizeof(UInt64);                        // size
     char buffer[buf_size];
     WriteBufferFromFileDescriptorDiscardOnFailure out(pipe.fds_rw[1], buf_size, buffer);