diff --git a/deps/oblib/src/lib/alloc/alloc_func.h b/deps/oblib/src/lib/alloc/alloc_func.h
index 85274c861a..36604b571c 100644
--- a/deps/oblib/src/lib/alloc/alloc_func.h
+++ b/deps/oblib/src/lib/alloc/alloc_func.h
@@ -48,6 +48,7 @@ int64_t ob_get_reserved_urgent_memory();
 // pc_pctg: percentage limitation of tenant memory can be used by Plan Cache
 // wa_pctg: percentage limitation of tenant memory can be used by Work Area
 
+int set_ctx_limit(uint64_t tenant_id, uint64_t ctx_id, const int64_t limit);
 int set_wa_limit(uint64_t tenand_id, int64_t wa_pctg);
 
 // set meta object memory limit for specified tenant.
diff --git a/mittest/mtlenv/test_multi_tenant.cpp b/mittest/mtlenv/test_multi_tenant.cpp
index 4375ad84cf..707229b46c 100644
--- a/mittest/mtlenv/test_multi_tenant.cpp
+++ b/mittest/mtlenv/test_multi_tenant.cpp
@@ -235,19 +235,21 @@ TEST_F(TestMultiTenant, get_tenant_context)
 class CtxMemConfigGetter : public ObICtxMemConfigGetter
 {
 public:
-  virtual int get(common::ObIArray<ObCtxMemConfig> &configs)
+  virtual int get(common::ObIArray<ObCtxMemConfig> &configs, int64_t tenant_limit)
   {
     int ret = OB_SUCCESS;
     {
       ObCtxMemConfig cfg;
       cfg.ctx_id_ = 10;
       cfg.idle_size_ = 10 * INTACT_ACHUNK_SIZE;
+      cfg.limit_ = 5 * tenant_limit / 100;
       ret = configs.push_back(cfg);
     }
     {
       ObCtxMemConfig cfg;
       cfg.ctx_id_ = 11;
       cfg.idle_size_ = 20 * INTACT_ACHUNK_SIZE;
+      cfg.limit_ = 10 * tenant_limit / 100;
       ret = configs.push_back(cfg);
     }
     return ret;
@@ -262,10 +264,11 @@ TEST_F(TestMultiTenant, idle)
   uint64_t tenant_id = 100;
   int ret = add_tenant(tenant_id);
   ASSERT_EQ(OB_SUCCESS, ret);
-  common::ObArray<ObCtxMemConfig> configs;
-  mcg.get(configs);
-  ASSERT_TRUE(2 == configs.size());
   ObMallocAllocator *malloc_allocator = ObMallocAllocator::get_instance();
+  int64_t tenant_limit = malloc_allocator->get_tenant_limit(tenant_id);
+  common::ObArray<ObCtxMemConfig> configs;
+  mcg.get(configs, tenant_limit);
+  ASSERT_TRUE(2 == configs.size());
   for (int i = 0 ; i < configs.size(); i++) {
     ObCtxMemConfig &cfg = configs.at(i);
     auto ta = malloc_allocator->get_tenant_ctx_allocator(tenant_id, cfg.ctx_id_);
@@ -273,6 +276,7 @@ TEST_F(TestMultiTenant, idle)
     int64_t chunk_cnt = cfg.idle_size_/INTACT_ACHUNK_SIZE;
     ASSERT_EQ(ta->chunk_cnt_, chunk_cnt);
     ASSERT_EQ(ta->idle_size_, cfg.idle_size_);
+    ASSERT_EQ(ta->get_limit(), cfg.limit_);
     ASSERT_NE(0, malloc_allocator->get_tenant_ctx_hold(tenant_id, cfg.ctx_id_));
   }
   // remove
diff --git a/src/observer/omt/ob_multi_tenant.cpp b/src/observer/omt/ob_multi_tenant.cpp
index 8a4d4fca36..bb5cbcc897 100644
--- a/src/observer/omt/ob_multi_tenant.cpp
+++ b/src/observer/omt/ob_multi_tenant.cpp
@@ -210,10 +210,17 @@ bool equal_with_tenant_id(const ObTenant *lhs,
   return NULL != lhs ? (lhs->id() == tenant_id) : false;
 }
 
-int ObCtxMemConfigGetter::get(common::ObIArray<ObCtxMemConfig> &configs)
+int ObCtxMemConfigGetter::get(common::ObIArray<ObCtxMemConfig> &configs, int64_t tenant_limit)
 {
-  UNUSED(configs);
-  return OB_SUCCESS;
+  int64_t ret = OB_SUCCESS;
+  {
+    ObCtxMemConfig cfg;
+    cfg.ctx_id_ = ObCtxIds::WORK_AREA;
+    cfg.idle_size_ = 0;
+    cfg.limit_ = 5 * tenant_limit / 100;
+    ret = configs.push_back(cfg);
+  }
+  return ret;
 }
 
 ObCtxMemConfigGetter g_default_mcg;
@@ -786,15 +793,18 @@ int ObMultiTenant::create_tenant(const ObTenantMeta &meta, bool write_slog, cons
   }
   if (OB_SUCC(ret)) {
     ObSEArray<ObCtxMemConfig, ObCtxIds::MAX_CTX_ID> configs;
-    if (OB_FAIL(mcg_->get(configs))) {
+    if (OB_FAIL(mcg_->get(configs, allowed_mem_limit))) {
       LOG_ERROR("get ctx mem config failed", K(ret));
     }
     for (int64_t i = 0; OB_SUCC(ret) && i < configs.count(); i++) {
       const uint64_t ctx_id = configs.at(i).ctx_id_;
       const int64_t idle_size = configs.at(i).idle_size_;
+      const int64_t limit = configs.at(i).limit_;
       const bool reserve = true;
       if (OB_FAIL(malloc_allocator->set_tenant_ctx_idle(tenant_id, ctx_id, idle_size, reserve))) {
         LOG_ERROR("set tenant ctx idle failed", K(ret));
+      } else if (OB_FAIL(set_ctx_limit(tenant_id, ctx_id, limit))) {
+        LOG_ERROR("set tenant ctx limit failed", K(ret), K(limit));
       }
       LOG_INFO("init ctx memory finish", K(ret), K(tenant_id), K(i), K(configs.at(i)));
     }
diff --git a/src/observer/omt/ob_multi_tenant.h b/src/observer/omt/ob_multi_tenant.h
index f1b427faed..33525b6c72 100644
--- a/src/observer/omt/ob_multi_tenant.h
+++ b/src/observer/omt/ob_multi_tenant.h
@@ -45,22 +45,23 @@ class ObTenantConfig;
 struct ObCtxMemConfig
 {
   ObCtxMemConfig()
-    : ctx_id_(0), idle_size_(0) {}
+    : ctx_id_(0), idle_size_(0), limit_(INT64_MAX) {}
   uint64_t ctx_id_;
   int64_t idle_size_;
-  TO_STRING_KV(K_(ctx_id), K_(idle_size));
+  int64_t limit_;
+  TO_STRING_KV(K_(ctx_id), K_(idle_size), K_(limit));
 };
 
 class ObICtxMemConfigGetter
 {
 public:
-  virtual int get(common::ObIArray<ObCtxMemConfig> &configs) = 0;
+  virtual int get(common::ObIArray<ObCtxMemConfig> &configs, int64_t tenant_limit) = 0;
 };
 
 class ObCtxMemConfigGetter : public ObICtxMemConfigGetter
 {
 public:
-  virtual int get(common::ObIArray<ObCtxMemConfig> &configs);
+  virtual int get(common::ObIArray<ObCtxMemConfig> &configs, int64_t tenant_limit);
 };
 
 // Forward declearation