Recommend the use of ES systems encountered in the process:
1. online business and offline business to be separated; (or otherwise write off a large number of polymerization investigation will lead to resource constraints (cpu idle reduced), leading online business slow query response)
2. control full gc avoid peak business execution;
3 data to be evenly distributed, can not be assured custom primary key is uniform, then the ES default primary key generation strategy;
same record 4 back and forth to modify the ES or concurrently inserted the same record, how do the records do not duplicate the problem: in a time less than index.refresh_interval offline Searching in side by side weight (remove duplicate, kept up to date)
5. to client.transport.sniff client is set to true