Terms table filled with duplicates and CPU thrash loop
When using an RSS feed connection as a post source with a hashtag applied, comments that include hashtags which are deleted will cause a runaway loop of creating new terms that match the Source Post's hashtag (I don't think the comment needs to match the hastag, any will do). This uses lots of CPU and creates thousands of terms entries.
I can supply an SQL backup of dev test instance ready to cause bug if needed.
do not do this on a used database.
- set users as Forum Mostly Public
- set admin/allow community tagging
- add rss feed as connection
- add New Source of feed with a tag eg "test"
- delete RSS connection, add it again to get a fresh version of the posts with tags (so we can comment on post, otherwise is locked?)
- click recent activity from connections for RSS connection
- click "view in context" on a post to see the comment field
- either make a comment with any hash tag (or maybe needs to match). then delete that comment with the hashtag. or add a tag, then delete the status
- check Database is looping. CPU use is 100%
small part of my log hubzillabuglog.txt
idea: one way to fix the table of any hubs which have had this happen (which could be any with rss and source!) is set UNIQUE contstrain using all fields except tid and term_hash, since you'd never need more than 1 of the same row anyway but I dunno if term_hash and parent_hash are used need expert here. ALTER IGNORE TABLE term ADD UNIQUE( aid, uid, oid, otype, ttype, term, url, imgurl, parent_hash);
The unique SQL constraint will fix databases but it wont stop the runaway process from happening.