[opt](memory) Modify memory gc conf and add `crash_in_alloc_large_mem…

…ory_bytes` (#39611)

1. faster frequency and small batch memory GC, which helps query
stability.
2. when alloc memory larger than crash_in_alloc_large_memory_bytes will
crash, default -1 means disabled. if you need a core dump to analyze
large memory allocation, modify this parameter to crash when large
memory allocation occur will help

be/src/common/config.cpp

@@ -125,8 +125,8 @@ DEFINE_Int64(max_sys_mem_available_low_water_mark_bytes, "6871947673");
 DEFINE_Int64(memtable_limiter_reserved_memory_bytes, "838860800");
 
 // The size of the memory that gc wants to release each time, as a percentage of the mem limit.
-DEFINE_mString(process_minor_gc_size, "10%");
-DEFINE_mString(process_full_gc_size, "20%");
+DEFINE_mString(process_minor_gc_size, "5%");
+DEFINE_mString(process_full_gc_size, "10%");
 
 // If true, when the process does not exceed the soft mem limit, the query memory will not be limited;
 // when the process memory exceeds the soft mem limit, the query with the largest ratio between the currently
@@ -140,6 +140,8 @@ DEFINE_mBool(enable_stacktrace, "true");
 
 DEFINE_mInt64(stacktrace_in_alloc_large_memory_bytes, "2147483648");
 
+DEFINE_mInt64(crash_in_alloc_large_memory_bytes, "-1");
+
 DEFINE_mBool(enable_memory_orphan_check, "false");
 
 // The maximum time a thread waits for full GC. Currently only query will wait for full gc.
@@ -588,7 +590,7 @@ DEFINE_mInt32(memory_maintenance_sleep_time_ms, "100");
 
 // After full gc, no longer full gc and minor gc during sleep.
 // After minor gc, no minor gc during sleep, but full gc is possible.
-DEFINE_mInt32(memory_gc_sleep_time_ms, "1000");
+DEFINE_mInt32(memory_gc_sleep_time_ms, "500");
 
 // Sleep time in milliseconds between memtbale flush mgr refresh iterations
 DEFINE_mInt64(memtable_mem_tracker_refresh_interval_ms, "5");

be/src/common/config.h

@@ -191,6 +191,10 @@ DECLARE_mBool(enable_stacktrace);
 // if alloc failed using Doris Allocator, will print stacktrace in error log.
 // if is -1, disable print stacktrace when alloc large memory.
 DECLARE_mInt64(stacktrace_in_alloc_large_memory_bytes);
+// when alloc memory larger than crash_in_alloc_large_memory_bytes will crash, default -1 means disabled.
+// if you need a core dump to analyze large memory allocation,
+// modify this parameter to crash when large memory allocation occur will help
+DECLARE_mInt64(crash_in_alloc_large_memory_bytes);
 
 // default is true. if any memory tracking in Orphan mem tracker will report error.
 DECLARE_mBool(enable_memory_orphan_check);

be/src/runtime/memory/thread_mem_tracker_mgr.h

@@ -242,17 +242,29 @@ inline void ThreadMemTrackerMgr::consume(int64_t size, int skip_large_memory_che
         flush_untracked_mem();
     }
 
-    if (skip_large_memory_check == 0 && doris::config::stacktrace_in_alloc_large_memory_bytes > 0 &&
-        size > doris::config::stacktrace_in_alloc_large_memory_bytes) {
-        _stop_consume = true;
-        LOG(WARNING) << fmt::format(
-                "malloc or new large memory: {}, {}, this is just a warning, not prevent memory "
-                "alloc, stacktrace:\n{}",
-                size,
-                is_attach_query() ? "in query or load: " + print_id(_query_id)
-                                  : "not in query or load",
-                get_stack_trace());
-        _stop_consume = false;
+    if (skip_large_memory_check == 0) {
+        if (doris::config::stacktrace_in_alloc_large_memory_bytes > 0 &&
+            size > doris::config::stacktrace_in_alloc_large_memory_bytes) {
+            _stop_consume = true;
+            LOG(WARNING) << fmt::format(
+                    "alloc large memory: {}, {}, this is just a warning, not prevent memory alloc, "
+                    "stacktrace:\n{}",
+                    size,
+                    is_attach_query() ? "in query or load: " + print_id(_query_id)
+                                      : "not in query or load",
+                    get_stack_trace());
+            _stop_consume = false;
+        }
+        if (doris::config::crash_in_alloc_large_memory_bytes > 0 &&
+            size > doris::config::crash_in_alloc_large_memory_bytes) {
+            LOG(FATAL) << fmt::format(
+                    "alloc large memory: {}, {}, crash generate core dumpsto help analyze, "
+                    "stacktrace:\n{}",
+                    size,
+                    is_attach_query() ? "in query or load: " + print_id(_query_id)
+                                      : "not in query or load",
+                    get_stack_trace());
+        }
     }
 }