I have a problem with PostgreSQL performance. I am using PostgreSQL 14.5 on Linux physical server. My query was taking 1 or 2 seconds yesterday, but today it is taking 5-6 seconds. Also, nothing has changed in the database.
Do you mean just no structural changes, or no changes of any kind?
Also, when I run the query for the second time, the query runs for the same time, so I expect it to be more performant the second time because of the cache.
(the database cache hit ratio is 97,8)
Lastly, some other queries take milliseconds normally, but sometimes they take minutes. I didn't understand why this is happening. (There is no lock, no table bloating and up to date analyze)
Did they start to "sometimes take minutes" just today, or was that a pre existing issue?
How up to date are the stats? Did you just finish ANALYZE right before you captured the plan?
This estimate looks pretty horrible and probably a problem with your stats:
-> Index Scan using "IX_T_WF_STEP_StepDefTypeCd" on "T_WF_STEP" wf_cstp (cost=0.43..1098968.54 rows=4019 width=8) (actual time=0.096..4278.668 rows=802420 loops=1)
And this one suggests you vacuuming is not very up to date:
Heap Fetches: 803045
I am adding the explain analyze output as an attachment.
Do you happen to have one for the same query from before the problem started? Also, it would be better to capture BUFFERS as part of the plan, and preferably with track_io_timing turned on.
