丢数据啊，丢数据

用datax倒cobar的数据时候，出现丢数据的情况。
我们现在5个分片。量级在2000多万。
走cobar的话，出现丢数据的情况，直接走mysql实例不存在丢的情况。
测试的时候发现，如果数据量过亿的话，倒到后面直接没反应了。貌似是超时断掉了。

Reproduce the transfer through Cobar with five shards, comparing it with a direct MySQL transfer at more than 20 million rows. Investigate the reported missing rows and the apparent timeout or hang beyond 100 million rows; done means transfers complete without loss and the results match the source.

Written by the indexing model from the issue text.