Feedback-ID: i76984098:Fastmail
Message-ID: <45cfb57b-9f34-4b83-a800-cc62542798ba@aklaver.com>
Date: Mon, 31 Mar 2025 08:08:00 -0700
MIME-Version: 1.0
User-Agent: Mozilla Thunderbird
Subject: Re: RDS IO Read time
To: Eden Aharoni <edena@legitsecurity.com>,
 "pgsql-general@lists.postgresql.org" <pgsql-general@lists.postgresql.org>
References: <DBBPR03MB102227BDDC40A11527BF6C0C0BBAD2@DBBPR03MB10222.eurprd03.prod.outlook.com>
Content-Language: en-US
From: Adrian Klaver <adrian.klaver@aklaver.com>
In-Reply-To: <DBBPR03MB102227BDDC40A11527BF6C0C0BBAD2@DBBPR03MB10222.eurprd03.prod.outlook.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
Archived-At: <https://www.postgresql.org/message-id/45cfb57b-9f34-4b83-a800-cc62542798ba%40aklaver.com>
Precedence: bulk

On 3/31/25 06:54, Eden Aharoni wrote:
> Hi all,
> 
> Hopefully, someone here could help us understand whats going on with our 
> deployment..
> 
> We are running Postgres 17.4 on AWS RDS on an ec2 instance that has:
> 
>   * 32vCPU
>   * 128GB RAM
>   * gp3 with 25K IOPS and 4000MiB/s throughput
>   * the instance supports up to 3125GB/s of throughput.
> 
> Whenever a query fetches data from the disk (index scan, bitmap scans, 
> etc.) we’re reaching 23 – 30MB/s of IO Read time.
> 
> We changed our autovacuum settings to be more aggressive (0 scale factor 
> and 10K threshold) so our index only scans will (hopefully) read data 
> from the disk rarely. However, we can’t optimize all the queries to use 
> index-only-scan since our users can dynamically select which columns to 
> see and filtering.
> 
> Is this expected IO read rate? I can’t help but feel we’re missing 
> something here..

RDS is a black box controlled by AWS, you are going to need to reach out 
to their tech support.

> 
> Thanks a lot! 😊
> 

-- 
Adrian Klaver
adrian.klaver@aklaver.com