很大可能是Bug，11.2.0.2 的expdp问题非常多，建议该版本不要用并行，或相同的测试在11.2.0.4上进行测试验证下效果。

类似于这个 Bug 18447188 - ORA-1 has occured after applying the fix of bug:18163767 (Doc ID 18447188.8) To BottomTo Bottom

Description
Exporting to an External Table can sometimes cause a deadlock.
One of such deadlock has been fixed by bug:18163767 so that deadlocked 
operation would be retried.  However, after applying the fix for bug:18163767 
export session may still fail due to unexpected ORA-1 error occurring during 
the retry of the operation after the ORA-60 deadlock. 
 
Rediscovery Notes
  
You might have encountered this bug if all of following conditions are true. 
 
- you are running multiple export (expdp) commands simultaneously 
- expdp uses external table method (access_method=external_table)
- patch to bug:18163767 has been applied
- export session is failing with unexpected ORA-1 error like the following:
 
 ------------------------------------------------------------------
 exported "TEST"."TEST_TBL22":"TEST_TBL22_P3" 139.4 KB  59 rows
 ORA-31693: "Table data object "TEST"."TEST_TBL22":"TEST_TBL22_P4" 
 failed to load/unload and is being skipped due to error:
 ORA-29913: error in executing ODCIEXTTABLEPOPULATE callout
 ORA-39068: invalid master table data in row with PROCESS_ORDER=4:1000001
 ORA-00001: unique constraint (SYSTEM.SYS_C001466840) violated
 ORA-06512: at "SYS.KUPF$FILE", line 3720
 ORA-06512: line 1
 ------------------------------------------------------------------
 
  
Workaround
  
Re-attemp export, or execute less number of exports at the same time.
重新尝试导出，或同时执行较少数量的导出。
复制

临时解决办法就是再次尝试导出或者不用并行。

使用parallel(expdp和impdp)参数来进行并行化。如果dumpfile指定一个文件，就没什么意义。在任意时刻只能有一个服务器进程进行写入操作，因此会存在一定的文件锁争用。最好是指定dumpfile=file%U.dmp ,每个进程写一个文件。如果你的存储io够快，那么导出导入的速度比较快，但是如果就一张表效果也并不是很好。