Re: survey: pg_stat_statements total_time and entry deallocation

On 10/5/18 13:47, Jeremy Schneider wrote:
> In addition to Kim's results and my initial data, a handful of people
> have forwarded results to me over the past few weeks.  Here's what came
> my way, aggregated and anonymized.

One more from a 9.4 system with default 5000 entries. Note that this is
a prod system but had only been gathering statements stats for about
three and a half days - the admin enabled pg_stat_statements last week
just to get stats for this email thread! (THANKS!!!)

While only one system in this thread has shown significant time that
would have rolled off, it is notable that the distribution of time
across buckets is always fairly lumpy. In fact I think that a good
number of systems have had more time in bucket 18 than bucket 19.

bucket entries max_calls total_time pct_time rows pct_rows
------- -------- ---------- ----------- --------- --------- ---------
1 248 1 1165381 0.24 9307526 3.67
2 248 1 278185 0.06 1947097 0.77
3 248 1 3223514 0.67 17897919 7.05
4 248 1 8722415 1.83 811022 0.32
5 248 1 413639 0.09 2103781 0.83
6 247 1 437788 0.09 2614929 1.03
7 247 1 202555 0.04 2094212 0.83
8 247 1 121673 0.03 350547 0.14
9 247 1 9299903 1.95 63600921 25.06
10 247 1 3642767 0.76 40022469 15.77
11 247 1 263853 0.06 2226582 0.88
12 247 1 4155721 0.87 14599659 5.75
13 247 1 1093557 0.23 16616897 6.55
14 247 1 19171119 4.01 686112 0.27
15 247 4 6929727 1.45 3761442 1.48
16 247 14 396828 0.08 73952 0.03
17 247 61 1888290 0.4 112077 0.04
18 247 326 131894870 27.6 5701711 2.25
19 247 1692 7856099 1.64 1529441 0.6
20 247 8120695 276724834 57.91 67741576 26.69

Thank you to everyone who is sending stats back! If anyone else out
there is reading this thread, it's never too late to add a little
additional data to the pile. :)

-Jeremy

--
Jeremy Schneider
Database Engineer
Amazon Web Services