From mboxrd@z Thu Jan 1 00:00:00 1970 From: Manavalan Krishnan Subject: Re: Large file copy to NFS mounted directory causes delay in other application packets Date: Wed, 9 Nov 2011 00:42:31 -0800 (PST) Message-ID: <1320828151.17101.YahooMailNeo@web160705.mail.bf1.yahoo.com> References: <1320816503.73813.YahooMailNeo@web160718.mail.bf1.yahoo.com> <1320819875.26025.48.camel@edumazet-laptop> Reply-To: Manavalan Krishnan Mime-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: netdev To: Dave Taht , Eric Dumazet , "linux-kernel@vger.kernel.org" Return-path: In-Reply-To: Sender: linux-kernel-owner@vger.kernel.org List-Id: netdev.vger.kernel.org >When I see behavior like this I keep thinking interactions between ove= rlarge large txqueuelens, somewhat busted TCP offloads on NICs, and tha= t pfifo_fast must die in favor of fair queuing and/or diffserv classifi= cation. But seeing >it on two different nics implies that your switch (= which I assume is shared) is possibly to blame... >(I see bufferbloat everywhere, but mostly because it's what I work on) >Is this NFS over TCP? Does the HA daemons prioritize packets at all? D= oes your switch? Does your qdiscs? How deep are your buffers on the net= work cards and txqueuelens and switch?=20 >(eric's other questions below are probably more valid) HA=A0 daemons does not prioritize. could you please provide info on ho= w to prioritize application packets? We tried with different switches, the same problem occurs. so switch ma= y not be the issue here. switch and qdisks does not prirotize the packets. Here are network buffers used in the servers txqueuelen is 1000=20 net.core.netdev_max_backlog =3D 1000 net.core.rmem_default =3D 262144 net.core.rmem_max =3D 262144 net.core.wmem_default =3D 129024 net.core.wmem_max =3D 131071 net.ipv4.tcp_rmem =3D 4096=A0=A0=A0 87380=A0=A0=A0 4194304 net.ipv4.tcp_wmem =3D 4096=A0=A0=A0 16384=A0=A0=A0 4194304 net.ipv4.tcp_mem =3D 196608=A0=A0=A0 262144=A0=A0=A0 393216 Thanks