Обсуждение: Understanding was terminated by signal 9: Killed

Wells Oliver <wells.oliver@gmail.com> writes:
> In an RDS instance with 16GB RAM, I ran a long query which started by
> setting temp_buffers to 16GB, so I think I plum ran out of memory, but can
> anyone point me in a different direction if the following log messages
> indicate something else is awry?

Yeah, this:

> 2021-10-19 21:10:37 UTC::@:[24752]:LOG:  server process (PID 25813) was
> terminated by signal 9: Killed

almost certainly indicates the Linux OOM killer at work.  If you were
running your own system I'd point you to [1], but I doubt that RDS
lets you put your hands on the relevant knobs.

            regards, tom lane

[1] https://www.postgresql.org/docs/current/kernel-resources.html#LINUX-MEMORY-OVERCOMMIT

You could always change the size of the RDS instance to something larger than what it currently is running, even just
temporarily.

Sent from my iPad

> On Oct 19, 2021, at 6:01 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>
> Wells Oliver <wells.oliver@gmail.com> writes:
>> In an RDS instance with 16GB RAM, I ran a long query which started by
>> setting temp_buffers to 16GB, so I think I plum ran out of memory, but can
>> anyone point me in a different direction if the following log messages
>> indicate something else is awry?
>
> Yeah, this:
>
>> 2021-10-19 21:10:37 UTC::@:[24752]:LOG:  server process (PID 25813) was
>> terminated by signal 9: Killed
>
> almost certainly indicates the Linux OOM killer at work.  If you were
> running your own system I'd point you to [1], but I doubt that RDS
> lets you put your hands on the relevant knobs.
>
>            regards, tom lane
>
> [1] https://www.postgresql.org/docs/current/kernel-resources.html#LINUX-MEMORY-OVERCOMMIT
>
>

Be careful!
I am not sure but upgrade for a larger instance can't be temporary. I think you can't downgrade to the original
instance class thereafter... 

--
Michel SALAIS

-----Message d'origine-----
De : John Scalia <jayknowsunix@gmail.com>
Envoyé : mercredi 20 octobre 2021 01:12
À : Tom Lane <tgl@sss.pgh.pa.us>
Cc : Wells Oliver <wells.oliver@gmail.com>; pgsql-admin <pgsql-admin@postgresql.org>
Objet : Re: Understanding was terminated by signal 9: Killed

You could always change the size of the RDS instance to something larger than what it currently is running, even just
temporarily.

Sent from my iPad

> On Oct 19, 2021, at 6:01 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>
> Wells Oliver <wells.oliver@gmail.com> writes:
>> In an RDS instance with 16GB RAM, I ran a long query which started by
>> setting temp_buffers to 16GB, so I think I plum ran out of memory,
>> but can anyone point me in a different direction if the following log
>> messages indicate something else is awry?
>
> Yeah, this:
>
>> 2021-10-19 21:10:37 UTC::@:[24752]:LOG:  server process (PID 25813)
>> was terminated by signal 9: Killed
>
> almost certainly indicates the Linux OOM killer at work.  If you were
> running your own system I'd point you to [1], but I doubt that RDS
> lets you put your hands on the relevant knobs.
>
>            regards, tom lane
>
> [1]
> https://www.postgresql.org/docs/current/kernel-resources.html#LINUX-ME
> MORY-OVERCOMMIT
>
>

That would be news to me. Where I currently work we’ve had to do that on several occasions as many of our dbs are in
themulti- multi- terabyte range, but some are just references without a lot of traffic. 

Sent from my iPad

> On Oct 20, 2021, at 4:27 PM, Michel SALAIS <msalais@msym.fr> wrote:
>
> Be careful!
> I am not sure but upgrade for a larger instance can't be temporary. I think you can't downgrade to the original
instance class thereafter... 
>
> --
> Michel SALAIS
>
> -----Message d'origine-----
> De : John Scalia <jayknowsunix@gmail.com>
> Envoyé : mercredi 20 octobre 2021 01:12
> À : Tom Lane <tgl@sss.pgh.pa.us>
> Cc : Wells Oliver <wells.oliver@gmail.com>; pgsql-admin <pgsql-admin@postgresql.org>
> Objet : Re: Understanding was terminated by signal 9: Killed
>
> You could always change the size of the RDS instance to something larger than what it currently is running, even just
temporarily.
>
> Sent from my iPad
>
>> On Oct 19, 2021, at 6:01 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>>
>> Wells Oliver <wells.oliver@gmail.com> writes:
>>> In an RDS instance with 16GB RAM, I ran a long query which started by
>>> setting temp_buffers to 16GB, so I think I plum ran out of memory,
>>> but can anyone point me in a different direction if the following log
>>> messages indicate something else is awry?
>>
>> Yeah, this:
>>
>>> 2021-10-19 21:10:37 UTC::@:[24752]:LOG:  server process (PID 25813)
>>> was terminated by signal 9: Killed
>>
>> almost certainly indicates the Linux OOM killer at work.  If you were
>> running your own system I'd point you to [1], but I doubt that RDS
>> lets you put your hands on the relevant knobs.
>>
>>           regards, tom lane
>>
>> [1]
>> https://www.postgresql.org/docs/current/kernel-resources.html#LINUX-ME
>> MORY-OVERCOMMIT
>>
>>
>
>

I'm 100% certain you can downsize given the selected instance type is compatible with your engine version.  Process is the same as scaling up and will result in downtime.  I don't see anything in documentation regarding AWS RDS Console, API's or Terraform documentation that would lead me to believe it cares about anything more than having valid parameters combinations regarless of engine / size.

Just to be certain, I created a t3.medium on my personal account with Postgres 13.3 and resized to a t3.micro immediately after it was provisioned.  No issues.

Summary of modifications

You are about to submit the following modifications. Only values that will change are displayed. Carefully verify your changes and click Modify DB Instance.

Attribute	Current value	New value
DB instance class	db.t3.medium	db.t3.micro

No issues Modifying.

Best regards,

Dan Smith

On Wed, Oct 20, 2021 at 1:27 PM Michel SALAIS <msalais@msym.fr> wrote:

Be careful!
I am not sure but upgrade for a larger instance can't be temporary. I think you can't downgrade to the original instance  class thereafter...

--
Michel SALAIS

-----Message d'origine-----
De : John Scalia <jayknowsunix@gmail.com>
Envoyé : mercredi 20 octobre 2021 01:12
À : Tom Lane <tgl@sss.pgh.pa.us>
Cc : Wells Oliver <wells.oliver@gmail.com>; pgsql-admin <pgsql-admin@postgresql.org>
Objet : Re: Understanding was terminated by signal 9: Killed

You could always change the size of the RDS instance to something larger than what it currently is running, even just temporarily.

Sent from my iPad

> On Oct 19, 2021, at 6:01 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>
> Wells Oliver <wells.oliver@gmail.com> writes:
>> In an RDS instance with 16GB RAM, I ran a long query which started by
>> setting temp_buffers to 16GB, so I think I plum ran out of memory,
>> but can anyone point me in a different direction if the following log
>> messages indicate something else is awry?
>
> Yeah, this:
>
>> 2021-10-19 21:10:37 UTC::@:[24752]:LOG:  server process (PID 25813)
>> was terminated by signal 9: Killed
>
> almost certainly indicates the Linux OOM killer at work.  If you were
> running your own system I'd point you to [1], but I doubt that RDS
> lets you put your hands on the relevant knobs.
>
>            regards, tom lane
>
> [1]
> https://www.postgresql.org/docs/current/kernel-resources.html#LINUX-ME
> MORY-OVERCOMMIT
>
>

--

Best regards,

Dan Smith