diff --git a/be/src/util/mem_info.h b/be/src/util/mem_info.h
index 9335933286ec24c..c17863cc56838f5 100644
--- a/be/src/util/mem_info.h
+++ b/be/src/util/mem_info.h
@@ -144,8 +144,11 @@ class MemInfo {
         if (config::enable_je_purge_dirty_pages) {
             try {
                 // Purge all unused dirty pages for arena <i>, or for all arenas if <i> equals MALLCTL_ARENAS_ALL.
-                jemallctl(fmt::format("arena.{}.purge", MALLCTL_ARENAS_ALL).c_str(), nullptr,
+                int err = jemallctl(fmt::format("arena.{}.purge", MALLCTL_ARENAS_ALL).c_str(), nullptr,
                           nullptr, nullptr, 0);
+                if (err) {
+                    LOG(WARNING) << "Jemalloc purge all unused dirty pages failed";
+                }
             } catch (...) {
                 LOG(WARNING) << "Purge all unused dirty pages for all arenas failed";
             }
@@ -153,6 +156,22 @@ class MemInfo {
 #endif
     }
 
+    // the limit of `tcache` is the number of pages, not the total number of page bytes.
+    // `tcache` has two cleaning opportunities: 1. the number of memory alloc and releases reaches a certain number,
+    // recycle pages that has not been used for a long time; 2. recycle all `tcache` when the thread exits.
+    // here add a total size limit.
+    static inline void je_thread_tcache_flush() {
+#ifdef USE_JEMALLOC
+        constexpr size_t TCACHE_LIMIT = (1ULL << 30); // 1G
+        if (allocator_cache_mem() - je_dirty_pages_mem() > TCACHE_LIMIT) {
+            int err = jemallctl("thread.tcache.flush", nullptr, nullptr, nullptr, 0);
+            if (err) {
+                LOG(WARNING) << "Jemalloc thread.tcache.flush failed";
+            }
+        }
+#endif
+    }
+
     static std::mutex je_purge_dirty_pages_lock;
     static std::condition_variable je_purge_dirty_pages_cv;
     static std::atomic<bool> je_purge_dirty_pages_notify;
diff --git a/be/src/vec/common/allocator.cpp b/be/src/vec/common/allocator.cpp
index 2b1c05533cd5049..a20882e3d05ad79 100644
--- a/be/src/vec/common/allocator.cpp
+++ b/be/src/vec/common/allocator.cpp
@@ -111,6 +111,8 @@ void Allocator<clear_memory_, mmap_populate, use_mmap>::sys_memory_check(size_t
                     print_id(doris::thread_context()->task_id()),
                     doris::thread_context()->get_thread_id(),
                     doris::config::thread_wait_gc_max_milliseconds, err_msg);
+            // only query thread exceeded memory limit for the first time and wait_gc is true.
+            doris::MemInfo::je_thread_tcache_flush();
             while (wait_milliseconds < doris::config::thread_wait_gc_max_milliseconds) {
                 std::this_thread::sleep_for(std::chrono::milliseconds(100));
                 if (!doris::GlobalMemoryArbitrator::is_exceed_hard_mem_limit(size)) {
diff --git a/bin/run-fs-benchmark.sh b/bin/run-fs-benchmark.sh
index 9908cc6c4d1da58..f4edd4117d01e8a 100755
--- a/bin/run-fs-benchmark.sh
+++ b/bin/run-fs-benchmark.sh
@@ -280,7 +280,7 @@ export LIBHDFS_OPTS="${final_java_opt}"
 #echo "LIBHDFS_OPTS: ${LIBHDFS_OPTS}"
 
 # see https://github.com/apache/doris/blob/master/docs/zh-CN/community/developer-guide/debug-tool.md#jemalloc-heap-profile
-export JEMALLOC_CONF="percpu_arena:percpu,background_thread:true,metadata_thp:auto,muzzy_decay_ms:15000,dirty_decay_ms:15000,oversize_threshold:0,prof:false,lg_prof_interval:32,lg_prof_sample:19,prof_gdump:false,prof_accum:false,prof_leak:false,prof_final:false"
+export JEMALLOC_CONF="percpu_arena:percpu,background_thread:true,metadata_thp:auto,muzzy_decay_ms:5000,dirty_decay_ms:5000,oversize_threshold:0,prof:false,lg_prof_interval:-1"
 export AWS_EC2_METADATA_DISABLED=true
 export AWS_MAX_ATTEMPTS=2
 
diff --git a/bin/start_be.sh b/bin/start_be.sh
index c2a4e5c8f9c828e..06a784bf2341a6a 100755
--- a/bin/start_be.sh
+++ b/bin/start_be.sh
@@ -403,7 +403,7 @@ export LIBHDFS_OPTS="${final_java_opt}"
 # log "LIBHDFS_OPTS: ${LIBHDFS_OPTS}"
 
 if [[ -z ${JEMALLOC_CONF} ]]; then
-    JEMALLOC_CONF="percpu_arena:percpu,background_thread:true,metadata_thp:auto,muzzy_decay_ms:15000,dirty_decay_ms:15000,oversize_threshold:0,prof:false,lg_prof_interval:32,lg_prof_sample:19,prof_gdump:false,prof_accum:false,prof_leak:false,prof_final:false"
+    JEMALLOC_CONF="percpu_arena:percpu,background_thread:true,metadata_thp:auto,muzzy_decay_ms:5000,dirty_decay_ms:5000,oversize_threshold:0,prof:false,lg_prof_interval:-1"
 fi
 
 if [[ -z ${JEMALLOC_PROF_PRFIX} ]]; then
diff --git a/cloud/script/start.sh b/cloud/script/start.sh
index e119e193d823f69..37d9abcb564e5a2 100644
--- a/cloud/script/start.sh
+++ b/cloud/script/start.sh
@@ -102,7 +102,7 @@ fi
 
 echo "LIBHDFS3_CONF=${LIBHDFS3_CONF}"
 
-export JEMALLOC_CONF="percpu_arena:percpu,background_thread:true,metadata_thp:auto,muzzy_decay_ms:15000,dirty_decay_ms:15000,oversize_threshold:0,prof:true,prof_prefix:jeprof.out"
+export JEMALLOC_CONF="percpu_arena:percpu,background_thread:true,metadata_thp:auto,muzzy_decay_ms:5000,dirty_decay_ms:5000,oversize_threshold:0,prof:false,lg_prof_interval:-1"
 
 mkdir -p "${DORIS_HOME}/log"
 echo "starts ${process} with args: $*"
diff --git a/conf/be.conf b/conf/be.conf
index 09b4d59daee3f1c..e3d952d9ffdfe7a 100644
--- a/conf/be.conf
+++ b/conf/be.conf
@@ -31,8 +31,8 @@ JAVA_OPTS_FOR_JDK_17="-Xmx1024m -DlogPath=$LOG_DIR/jni.log -Xlog:gc*:$LOG_DIR/be
 
 # https://github.com/apache/doris/blob/master/docs/zh-CN/community/developer-guide/debug-tool.md#jemalloc-heap-profile
 # https://jemalloc.net/jemalloc.3.html
-JEMALLOC_CONF="percpu_arena:percpu,background_thread:true,metadata_thp:auto,muzzy_decay_ms:15000,dirty_decay_ms:15000,oversize_threshold:0,prof:false,lg_prof_interval:32,lg_prof_sample:19,prof_gdump:false,prof_accum:false,prof_leak:false,prof_final:false"
-JEMALLOC_PROF_PRFIX=""
+JEMALLOC_CONF="percpu_arena:percpu,background_thread:true,metadata_thp:auto,muzzy_decay_ms:5000,dirty_decay_ms:5000,oversize_threshold:0,prof:false,lg_prof_interval:-1"
+JEMALLOC_PROF_PRFIX="jemalloc_heap_profile_"
 
 # ports for admin, web, heartbeat service 
 be_port = 9060
diff --git a/regression-test/pipeline/external/conf/be.conf b/regression-test/pipeline/external/conf/be.conf
index 94a038cfa885fb2..306459336222c61 100644
--- a/regression-test/pipeline/external/conf/be.conf
+++ b/regression-test/pipeline/external/conf/be.conf
@@ -30,8 +30,8 @@ JAVA_OPTS_FOR_JDK_17="-Xmx1024m -DlogPath=$DORIS_HOME/log/jni.log -Xlog:gc*:$DOR
 
 # https://github.com/apache/doris/blob/master/docs/zh-CN/community/developer-guide/debug-tool.md#jemalloc-heap-profile
 # https://jemalloc.net/jemalloc.3.html
-JEMALLOC_CONF="percpu_arena:percpu,background_thread:true,metadata_thp:auto,muzzy_decay_ms:15000,dirty_decay_ms:15000,oversize_threshold:0,prof:false,lg_prof_interval:32,lg_prof_sample:19,prof_gdump:false,prof_accum:false,prof_leak:false,prof_final:false"
-JEMALLOC_PROF_PRFIX=""
+JEMALLOC_CONF="percpu_arena:percpu,background_thread:true,metadata_thp:auto,muzzy_decay_ms:5000,dirty_decay_ms:5000,oversize_threshold:0,prof:false,lg_prof_interval:-1"
+JEMALLOC_PROF_PRFIX="jemalloc_heap_profile_"
 
 # INFO, WARNING, ERROR, FATAL
 sys_log_level = INFO
diff --git a/regression-test/pipeline/p0/conf/be.conf b/regression-test/pipeline/p0/conf/be.conf
index a072ac7ad50aca4..e72709603d77cd4 100644
--- a/regression-test/pipeline/p0/conf/be.conf
+++ b/regression-test/pipeline/p0/conf/be.conf
@@ -30,8 +30,8 @@ JAVA_OPTS_FOR_JDK_17="-Xmx1024m -DlogPath=$DORIS_HOME/log/jni.log -Xlog:gc*:$DOR
 
 # https://github.com/apache/doris/blob/master/docs/zh-CN/community/developer-guide/debug-tool.md#jemalloc-heap-profile
 # https://jemalloc.net/jemalloc.3.html
-JEMALLOC_CONF="percpu_arena:percpu,background_thread:true,metadata_thp:auto,muzzy_decay_ms:15000,dirty_decay_ms:15000,oversize_threshold:0,prof:false,lg_prof_interval:32,lg_prof_sample:19,prof_gdump:false,prof_accum:false,prof_leak:false,prof_final:false"
-JEMALLOC_PROF_PRFIX=""
+JEMALLOC_CONF="percpu_arena:percpu,background_thread:true,metadata_thp:auto,muzzy_decay_ms:5000,dirty_decay_ms:5000,oversize_threshold:0,prof:false,lg_prof_interval:-1"
+JEMALLOC_PROF_PRFIX="jemalloc_heap_profile_"
 
 # INFO, WARNING, ERROR, FATAL
 sys_log_level = INFO
diff --git a/regression-test/pipeline/p1/conf/be.conf b/regression-test/pipeline/p1/conf/be.conf
index 675518ac0ce464a..466f0d589614270 100644
--- a/regression-test/pipeline/p1/conf/be.conf
+++ b/regression-test/pipeline/p1/conf/be.conf
@@ -30,8 +30,8 @@ JAVA_OPTS_FOR_JDK_17="-Xmx1024m -DlogPath=$DORIS_HOME/log/jni.log -Xlog:gc*:$DOR
 
 # https://github.com/apache/doris/blob/master/docs/zh-CN/community/developer-guide/debug-tool.md#jemalloc-heap-profile
 # https://jemalloc.net/jemalloc.3.html
-JEMALLOC_CONF="percpu_arena:percpu,background_thread:true,metadata_thp:auto,muzzy_decay_ms:15000,dirty_decay_ms:15000,oversize_threshold:0,prof:false,lg_prof_interval:32,lg_prof_sample:19,prof_gdump:false,prof_accum:false,prof_leak:false,prof_final:false"
-JEMALLOC_PROF_PRFIX=""
+JEMALLOC_CONF="percpu_arena:percpu,background_thread:true,metadata_thp:auto,muzzy_decay_ms:5000,dirty_decay_ms:5000,oversize_threshold:0,prof:false,lg_prof_interval:-1"
+JEMALLOC_PROF_PRFIX="jemalloc_heap_profile_"
 
 # INFO, WARNING, ERROR, FATAL
 sys_log_level = INFO