MIME-Version: 1.0
References: 
 <BL0PR06MB5009A7B383ACA34CD3603B058A82A@BL0PR06MB5009.namprd06.prod.outlook.com>
In-Reply-To: 
 <BL0PR06MB5009A7B383ACA34CD3603B058A82A@BL0PR06MB5009.namprd06.prod.outlook.com>
From: Joshua Banton <bantonj@gmail.com>
Date: Fri, 9 Jan 2026 16:58:57 -0500
Message-ID: 
 <CAHkYM9bP5izAeh_JF1PqM6v4HVfQOgeEmCQF0i63oJFFwnUsdA@mail.gmail.com>
Subject: Re: Troubleshooting a brief spike in active sessions
To: "Dirschel, Steve-CW" <Steve.Dirschel@bestbuy.com>
Cc: "pgsql-performance@lists.postgresql.org"
 <pgsql-performance@lists.postgresql.org>
Content-Type: multipart/alternative; boundary="000000000000759cad0647fba20b"
Archived-At: 
 <https://www.postgresql.org/message-id/CAHkYM9bP5izAeh_JF1PqM6v4HVfQOgeEmCQF0i63oJFFwnUsdA%40mail.gmail.com>
Precedence: bulk

--000000000000759cad0647fba20b
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

On Fri, Jan 9, 2026 at 4:41=E2=80=AFPM Dirschel, Steve-CW <
Steve.Dirschel@bestbuy.com> wrote:

> Aurora Postgres version 17.4
>
> This application mainly runs 3 queries very frequently-  around 12000
> executions per second across the 3.  Normally the server handles this fin=
e
> and at any given time there are between 0 - 10 active sessions running th=
e
> queries.  The 3 queries are tuned-  doing 3-7 logical reads per execute a=
nd
> run in < .1 millisecond.
>
> Periodically there will be a huge spike in active sessions up to 400+
> sessions all on CPU (it can be higher depending on how many app servers a=
re
> in the mix at the time).  This spike will last 2-5 seconds but it causes =
an
> increase in app response time which is a problem.
>

I'm not super familiar with Aurora, but I've seen this caused by a few
different things in vanilla Postgres. I don't think Aurora has checkpoints,
so it's probably not that. I have seen where a vacuum can cause various
Postgres buffers to have a lot of churn, which can cause higher query
latency for a period of time. I have also seen temporary increased disk
latency to cause higher query latency.

Josh B

--000000000000759cad0647fba20b
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><div dir=3D"ltr"><br></div><br><div class=3D"gmail_quote g=
mail_quote_container"><div dir=3D"ltr" class=3D"gmail_attr">On Fri, Jan 9, =
2026 at 4:41=E2=80=AFPM Dirschel, Steve-CW &lt;<a href=3D"mailto:Steve.Dirs=
chel@bestbuy.com">Steve.Dirschel@bestbuy.com</a>&gt; wrote:<br></div><block=
quote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1=
px solid rgb(204,204,204);padding-left:1ex">


<div>
<div dir=3D"ltr" style=3D"font-family:Aptos,Arial,Helvetica,sans-serif;font=
-size:12pt;color:rgb(0,0,0)">
Aurora Postgres version 17.4</div>
<div dir=3D"ltr" style=3D"font-family:Aptos,Arial,Helvetica,sans-serif;font=
-size:12pt;color:rgb(0,0,0)">
<br>
</div>
<div dir=3D"ltr" style=3D"font-family:Aptos,Arial,Helvetica,sans-serif;font=
-size:12pt;color:rgb(0,0,0)">
This application mainly runs 3 queries very frequently- =C2=A0around 12000 =
executions per second across the 3.=C2=A0 Normally the server handles this =
fine and at any given time there are between 0 - 10 active sessions running=
 the queries.=C2=A0 The 3 queries are tuned- =C2=A0doing
 3-7 logical reads per execute and run in &lt; .1 millisecond.</div>
<div dir=3D"ltr" style=3D"font-family:Aptos,Arial,Helvetica,sans-serif;font=
-size:12pt;color:rgb(0,0,0)">
<br>
</div>
<div dir=3D"ltr" style=3D"font-family:Aptos,Arial,Helvetica,sans-serif;font=
-size:12pt;color:rgb(0,0,0)">
Periodically there will be a huge spike in active sessions up to 400+ sessi=
ons all on CPU (it can be higher depending on how many app servers are in t=
he mix at the time).=C2=A0 This spike will last 2-5 seconds but it causes a=
n increase in app response time which
 is a problem.=C2=A0=C2=A0</div></div></blockquote><div>=C2=A0</div><font s=
ize=3D"2">I&#39;m not super familiar with Aurora, but I&#39;ve seen this ca=
used by a few different things in vanilla Postgres. I don&#39;t think Auror=
a has checkpoints, so it&#39;s probably not that. I have seen where a vacuu=
m can cause various Postgres buffers to have a lot of churn, which can caus=
e higher query latency for a period of time. I have also seen temporary inc=
reased disk latency to cause higher query latency.</font></div><div class=
=3D"gmail_quote gmail_quote_container"><font size=3D"2"><br></font></div><d=
iv class=3D"gmail_quote gmail_quote_container"><font size=3D"2">Josh B</fon=
t></div></div>

--000000000000759cad0647fba20b--