MIME-Version: 1.0
References: 
 <BL0PR06MB5009A7B383ACA34CD3603B058A82A@BL0PR06MB5009.namprd06.prod.outlook.com>
In-Reply-To: 
 <BL0PR06MB5009A7B383ACA34CD3603B058A82A@BL0PR06MB5009.namprd06.prod.outlook.com>
From: Nikolay Samokhvalov <nik@postgres.ai>
Date: Fri, 9 Jan 2026 14:27:17 -0800
Message-ID: 
 <CAM527d-Udxg5NSg_kEsEuTm-gB5GJ8z=84znpTQQT-tu2VQW_A@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="000000000000cb84c20647fc0745"
Archived-At: 
 <https://www.postgresql.org/message-id/CAM527d-Udxg5NSg_kEsEuTm-gB5GJ8z%3D84znpTQQT-tu2VQW_A%40mail.gmail.com>
Precedence: bulk

--000000000000cb84c20647fc0745
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

On Fri, Jan 9, 2026 at 1:41=E2=80=AFPM Dirschel, Steve-CW <
Steve.Dirschel@bestbuy.com> wrote:
>
> Aurora Postgres version 17.4
>
> ...
>
> The app sessions will show as wait_event_type and wait_event as null
which I understand means they are on CPU
>
> ...


Not necessarily. It also might be an uninstrumented wait event.

Do the spikes of "CPU" AAS (I prefer to call them "CPU*") correspond well
to the spikes of high usage?

If you have N active sessions marked as CPU in ASH, does it match to the
CPU load of similar scale?

If the VM has much less than 400 vCPUs and you see overall CPU load far
from 100%, then we should suspect some missing wait event.

Nik

--000000000000cb84c20647fc0745
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr">On Fri, Jan 9, 2026 at 1:41=E2=80=AFPM Dirschel, Steve-CW =
&lt;<a href=3D"mailto:Steve.Dirschel@bestbuy.com">Steve.Dirschel@bestbuy.co=
m</a>&gt; wrote:<br>&gt;<br>&gt; Aurora Postgres version 17.4<br>&gt;<br>&g=
t; ...<br>&gt;<br>&gt; The app sessions will show as wait_event_type and wa=
it_event as null which I understand means they are on CPU<br>&gt;<br>&gt; .=
..<br><br><br>Not necessarily. It also might be an uninstrumented wait even=
t.<br><br>Do the spikes of &quot;CPU&quot; AAS (I prefer to call them &quot=
;CPU*&quot;) correspond well to the spikes of high usage?<br><br>If you hav=
e N active sessions marked as CPU in ASH, does it match to the CPU load of =
similar scale?<br><br>If the VM has much less than 400 vCPUs and you see ov=
erall CPU load far from 100%, then we should suspect some missing wait even=
t.<br><br>Nik</div>

--000000000000cb84c20647fc0745--