In my case I drop a few joins, deploy it, but join on other table, that I not even touched, started use seq scan instead of index. I saw x10 query degradation and rollback release.
After analyze query plans started use index again.
Top-level
In my case I drop a few joins, deploy it, but join on other table, that I not even touched, started use seq scan instead of index. I saw x10 query degradation and rollback release. After analyze query plans started use index again. 1 comment
|
But after that I spotted a few more degradated queries. And rolled back release again. But after research I figured out that queries always have pikes on latency graphic at this time of day.
So the point is: keep calm and profile query behaviour not only for few hours, but for a couple days too.