Received: from malur.postgresql.org ([217.196.149.56]) by arkaria.postgresql.org with esmtps (TLS1.3) tls TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384 (Exim 4.94.2) (envelope-from ) id 1tzJAw-000bc9-UJ for pgsql-general@arkaria.postgresql.org; Mon, 31 Mar 2025 17:43:54 +0000 Received: from localhost ([127.0.0.1] helo=malur.postgresql.org) by malur.postgresql.org with esmtp (Exim 4.94.2) (envelope-from ) id 1tzJAv-009ZZj-LG for pgsql-general@arkaria.postgresql.org; Mon, 31 Mar 2025 17:43:53 +0000 Received: from makus.postgresql.org ([2001:4800:3e1:1::229]) by malur.postgresql.org with esmtps (TLS1.3) tls TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384 (Exim 4.94.2) (envelope-from ) id 1tzJAv-009ZZb-A5 for pgsql-general@lists.postgresql.org; Mon, 31 Mar 2025 17:43:53 +0000 Received: from smtp115.iad3b.emailsrvr.com ([146.20.161.115]) by makus.postgresql.org with esmtps (TLS1.2) tls TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384 (Exim 4.96) (envelope-from ) id 1tzJAt-002BKm-1e for pgsql-general@lists.postgresql.org; Mon, 31 Mar 2025 17:43:52 +0000 X-Auth-ID: xof@thebuild.com Received: by smtp15.relay.iad3b.emailsrvr.com (Authenticated sender: xof-AT-thebuild.com) with ESMTPSA id 2B749C00CA; Mon, 31 Mar 2025 13:43:50 -0400 (EDT) Content-Type: text/plain; charset=us-ascii Mime-Version: 1.0 (Mac OS X Mail 16.0 \(3776.700.51.11.1\)) Subject: Re: [EXTERNAL] RDS IO Read time From: Christophe Pettus In-Reply-To: Date: Mon, 31 Mar 2025 10:43:19 -0700 Cc: "pgsql-general@lists.postgresql.org" Content-Transfer-Encoding: quoted-printable Message-Id: References: To: Eden Aharoni X-Mailer: Apple Mail (2.3776.700.51.11.1) X-Classification-ID: 7ff1c816-90d5-4475-8f10-01af8564772b-1-1 List-Id: List-Help: List-Subscribe: List-Post: List-Owner: List-Archive: Archived-At: Precedence: bulk > On Mar 31, 2025, at 10:32, Eden Aharoni = wrote: >=20 > First, thanks for the reply :) > So, I do know which part is taking a lot of IO time and it's to be = honest any node that reads from the disk.. of course, we're running = EXPLAIN on our queries (to be more specific we use auto_explain) but we = can't seem to find what could cause an index scan that reads 34 MB to = take more than a second (1.2 sec).. we do know that our dataset doesn't = fit the memory and we were ok with that but with IO being so slow we = just don't know what to do anymore. Any other suggestions other than = contact AWS (which we did but it seems this path won't lead anywhere). The performance of EBS is definitely a question for AWS. I can say = that, in our experience, EBS mounts for RDS almost never approach the = stated maximum throughput, although io2 tends to be closer than gp2 or = gp3.=