Operating System: windows2012
Database version: Version 11.2.0.1 local database dblink connection is 12.1.0.2 (aws cloud)
problem:
The company's backup is expdp backup timing software, local windows oracle database by db_link backup, which by backup software has an error, which perform expdp statement alone there are error, ORA-31626: job does not exist ORA-31650: timeout waiting for master process response
Screenshot below:
Meanwhile, the company's software task scheduling regular backups shots :( The following diagram is a screenshot below has been handled well, forget the previous screenshot)
Description:
Above screenshot, the situation is rds_pay regular backup task fails, not only is the timing task fails, the failure to execute the script alone!
ORA-31626: job does not exist ORA-31650: timeout waiting for master process response
Roadmap:
Since a separate window in cmd Run are given (with the db_link), used in the machine by plsql query no problem db_link
Description dblink is not accessible through the problem, that the network is no problem, no problem dblink, and 12c library is no problem, the problem on the local library 11G, the following are found through the support of several situations:
1.aq_tm_processes parameters
10g This parameter sets the range 0-10,11g is 0-40, then the local database is set 4, I set to 16, Oracle does not recommend the official set to 0, it can not be monitored; revised to 16, to take effect immediately, or the same error, so obviously this is not a small problem parameters.
2.升级后的问题(这个排除，没有进行升级不一致产生的，需要重新执行catalog.sql and catproc.sql来配置后)
3.rac数据库没有设置clsuter=no （很明显不是rac数据库，排除）
4. SGA设置可能偏小
查看了support文档,有提到至少sga为1G，但是结合现实中的情况，有接近10个定时任务的窗口，每个任务1G+，那么就定时任务耗费的SGA就有10G+；
整个memory_max_target 为17G，明显偏小（10个备份定时任务），在设置本地11g数据库memory_max_target为30g（服务器内存64G,先试一试30G，最大设置成物理内存的80%）,重启后，就能够正常备份，问题得到解决。
5.触发未知的bug(11.2.0.1通过DBLINK导出的12.1.0.2.0数据库数据)
既然4解决了，那就不是BUG了。
总结：
本次出现ORA-31626: job does not exist ORA-31650: timeout waiting for master process response报错的原因是SGA过小，虽然support文档有提到至少1G，但是这个只是一个思路，并不是真的只有1G就够了，但是我也是想了很久才打算调整内存参数的；
顺便说一下如果有提示 ORA-31634: job already exists;sys_export_开头的表太多，需要清理：
SELECT owner_name, job_name, rtrim(operation) "OPERATION",
rtrim(job_mode) "JOB_MODE", state, attached_sessions
FROM dba_datapump_jobs
WHERE job_name NOT LIKE 'BIN$%'
ORDER BY 1,2;
其中state为not running状态的直接删除就行，executing状态的表不能够删除。删除后就可以备份了。