We have a website and its database is running on the AWS RDS.
There is only one FULLTEXT index on a innodb table for the site search feature.
Most of our data contains CJK characters, due to the behavior of default parser is not suitable for CJK characters even though we set innodb_ft_min_token_size as 1, we want to change to use N-Gram parser.
With N-Gram parser, the FULLTEXT search results are better than previous. However, the CPU utilization is continuously at high value (over 70%), and very easy to cause the RDS server no response. Here is one of example images.
Moreover, sometimes it will keep stuck at 99% usage and very quickly cause the RDS server down.
I have used show processlist to check the running threads, but the traffic of our site is light. There are only a few queries, and also which are not complicated.
If we build FULLTEXT index without N-Gram parser, the CPU utilization is stably at the range from 5% to 20%. All thing will work fine, only the search results are not satisfactory.
Are there any solutions for this problem?
