We're beginning to see a lot of slow queries with this table.
It has about 28 million rows in it. It appears it stores word-post_id-title info. Is there a way to purge the table or optimize the feature? [originally posted on jforum.net by kkim]
I'm a sys admin/dba and we're exploring ways to spare some of the load from the database server. It looks like when search indexing is enabled, it generates a heavy load ofselects/inserts to the search tables on each post operation. With it disabled, the savings are drastic.
My question is -
Does it effectively disable the search feature if we set
search.indexing.enabled = false
or is there an alternative? It seems like there's an option to run the indexer as a scheduled background task but I can't tell if that can't be activated with the above option off.
Ah, yes, the indexing part of the search engine will parse each post into words, then do a search for existing words in the words table. But I think most DB implementations try to do this via an "in clause". However, each new word in a post will generate an insert into the words table and each word in a post will generate an insert into the words to post table.
From what I can tell from the code, to change from indexing at post time to indexing on a schedule, you need to change the search.indexer.implementation to the following: