你好，
我们运行的是基于 Docker 的 Discourse 2.7.7，拥有庞大的用户群。从 2.4.2 升级到 2.7.7 后，我们发现部分 Postgres 进程 CPU 占用率持续处于 100%。进一步调查表明，以下查询似乎是导致问题的原因：
discourse_prod=# select pid, datid, query from pg_stat_activity WHERE pid = '244906';
pid | datid | query
--------+---------+-------------------------------------------------------
244906 | 2068583 | DELETE FROM user_badges +
| | WHERE id IN ( +
| | SELECT ub.id +
| | FROM user_badges ub +
| | LEFT JOIN ( +
| | SELECT id user_id, current_timestamp granted_at+
| | FROM users +
| | WHERE id IN ( +
| | SELECT p1.user_id +
| | FROM post_custom_fields pc +
| | JOIN badge_posts p1 ON p1.id = pc.post_id +
| | JOIN topics t1 ON p1.topic_id = t1.id +
| | WHERE p1.user_id <> t1.user_id AND +
| | name = 'is_accepted_answer' AND +
| | p1.user_id IN ( +
| | SELECT user_id +
| | FROM posts +
| | WHERE TRUE OR p1.id IN (-1) +
| | ) +
| | GROUP BY p1.user_id +
| | HAVING COUNT(*) > 9 +
| | ) +
| | ) q ON q.user_id = ub.user_id +
| | +
| | WHERE ub.badge_id = 103 AND q.user_id IS NULL +
| | ) +
| |
(1 row)
我们发现这个查询，我推测它会导致巨大的开销，对吧？
| | SELECT user_id +
| | FROM posts +
| | WHERE TRUE OR p1.id IN (-1) +
对此有什么建议吗？
如果需要更多信息，请告诉我。
谢谢，
Daniel。