diff --git a/drape/batcher.cpp b/drape/batcher.cpp
index 4cb6f9db63..c38cc5c45e 100644
--- a/drape/batcher.cpp
+++ b/drape/batcher.cpp
@@ -177,15 +177,20 @@ RefPointer<RenderBucket> Batcher::GetBucket(GLState const & state)
 void Batcher::FinalizeBucket(GLState const & state)
 {
   ASSERT(m_buckets.find(state) != m_buckets.end(), ("Have no bucket for finalize with given state"));
-  m_flushInterface(state, m_buckets[state].Move());
+  MasterPointer<RenderBucket> bucket = m_buckets[state];
   m_buckets.erase(state);
+  bucket->GetBuffer()->Preflush();
+  m_flushInterface(state, bucket.Move());
 }
 
 void Batcher::Flush()
 {
   ASSERT(m_flushInterface != NULL, ());
   for (buckets_t::iterator it = m_buckets.begin(); it != m_buckets.end(); ++it)
+  {
+    it->second->GetBuffer()->Preflush();
     m_flushInterface(it->first, it->second.Move());
+  }
 
   m_buckets.clear();
 }
diff --git a/drape/glfunctions.cpp b/drape/glfunctions.cpp
index 9d8a2729c2..6f58c24ea9 100644
--- a/drape/glfunctions.cpp
+++ b/drape/glfunctions.cpp
@@ -4,6 +4,12 @@
 #include "../base/assert.hpp"
 #include "../base/logging.hpp"
 
+#ifdef DEBUG
+#include "../base/thread.hpp"
+#include "../base/mutex.hpp"
+#include "../std/map.hpp"
+#endif
+
 #include "../std/cstring.hpp"
 
 #ifndef OMIM_OS_WINDOWS
@@ -12,6 +18,14 @@
 
 namespace
 {
+#ifdef DEBUG
+  typedef pair<threads::ThreadID, glConst> TKey;
+  typedef pair<TKey, uint32_t> TNode;
+  typedef map<TKey, uint32_t> TBoundMap;
+  TBoundMap g_boundBuffers;
+  threads::Mutex g_mutex;
+#endif
+
   inline GLboolean convert(bool v)
   {
     return (v == true) ? GL_TRUE : GL_FALSE;
@@ -294,12 +308,26 @@ uint32_t GLFunctions::glGenBuffer()
 void GLFunctions::glBindBuffer(uint32_t vbo, uint32_t target)
 {
   ASSERT(glBindBufferFn != NULL, ());
+#ifdef DEBUG
+  threads::MutexGuard guard(g_mutex);
+  TKey key = make_pair(threads::GetCurrentThreadID(), target);
+  auto iter = g_boundBuffers.find(key);
+  if (iter != g_boundBuffers.end())
+    g_boundBuffers.erase(iter);
+  g_boundBuffers.emplace(key, vbo);
+#endif
   GLCHECK(glBindBufferFn(target, vbo));
 }
 
 void GLFunctions::glDeleteBuffer(uint32_t vbo)
 {
   ASSERT(glDeleteBuffersFn != NULL, ());
+#ifdef DEBUG
+  threads::MutexGuard guard(g_mutex);
+  threads::ThreadID id = threads::GetCurrentThreadID();
+  for (TNode const & n : g_boundBuffers)
+    ASSERT(n.second != vbo, ());
+#endif
   GLCHECK(glDeleteBuffersFn(1, &vbo));
 }
 
@@ -615,6 +643,9 @@ void GLFunctions::glDrawElements(uint16_t indexCount)
 void CheckGLError()
 {
   GLenum result = glGetError();
-  if (result != GL_NO_ERROR)
+  while (result != GL_NO_ERROR)
+  {
     LOG(LERROR, ("GLError:", result));
+    result = glGetError();
+  }
 }
diff --git a/drape/gpu_buffer.cpp b/drape/gpu_buffer.cpp
index 6c76f52d4d..8a9e2674e3 100644
--- a/drape/gpu_buffer.cpp
+++ b/drape/gpu_buffer.cpp
@@ -39,6 +39,7 @@ GPUBuffer::GPUBuffer(Target t, uint8_t elementSize, uint16_t capacity)
 
 GPUBuffer::~GPUBuffer()
 {
+  GLFunctions::glBindBuffer(0, glTarget(m_t));
   GLFunctions::glDeleteBuffer(m_bufferID);
 }
 
@@ -56,7 +57,7 @@ void GPUBuffer::UploadData(void const * data, uint16_t elementCount)
 
 void GPUBuffer::Bind()
 {
-  GLFunctions::glBindBuffer(m_bufferID, glTarget((m_t)));
+  GLFunctions::glBindBuffer(m_bufferID, glTarget(m_t));
 }
 
 void * GPUBuffer::Map()
diff --git a/drape/vertex_array_buffer.cpp b/drape/vertex_array_buffer.cpp
index b041a216a4..baa3b0d5fc 100644
--- a/drape/vertex_array_buffer.cpp
+++ b/drape/vertex_array_buffer.cpp
@@ -32,6 +32,12 @@ VertexArrayBuffer::~VertexArrayBuffer()
   }
 }
 
+void VertexArrayBuffer::Preflush()
+{
+  GLFunctions::glBindBuffer(0, gl_const::GLElementArrayBuffer);
+  GLFunctions::glBindBuffer(0, gl_const::GLArrayBuffer);
+}
+
 void VertexArrayBuffer::Render()
 {
   if (!(m_staticBuffers.empty() && m_dynamicBuffers.empty()))
diff --git a/drape/vertex_array_buffer.hpp b/drape/vertex_array_buffer.hpp
index 855ff32d56..0b2d80b101 100644
--- a/drape/vertex_array_buffer.hpp
+++ b/drape/vertex_array_buffer.hpp
@@ -20,6 +20,9 @@ public:
   VertexArrayBuffer(uint32_t indexBufferSize, uint32_t dataBufferSize);
   ~VertexArrayBuffer();
 
+  /// This method must be call on reading thread, before VAO will be transfer on render thread
+  void Preflush();
+
   ///{@
   /// On devices where implemented OES_vertex_array_object extensions we use it for build VertexArrayBuffer
   /// OES_vertex_array_object create OpenGL resource that belong only one GL context (which was created by)