Hello, I am currently experiencing an issue with FreeBSD 9.0-RC2 r227852 where the net.inet.tcp.reass.curesegments value is constantly increasing (and not descreasing when there is nominal traffic with the box). It is causing tcp slowdowns as described with kern/155407: Exhausted net.inet.tcp.reass.maxsegments block recovering tcp session (for this socket and any other socket waiting for retransmited packets). After exhausted net.inet.tcp.reass.maxsegments allocation new entry in tcp_reass failed (for this socket and any other socket waiting for retransmited packets). I have increased the reass.maxsegments value to 16384 to temporarily avoid the problem, but the cursegments number keeps rising and it seems it will occur again. Is this an issue that anyone else has seen? I can provide more information if need be. Thanks, Kris
Hi, I 'm experiencing a similar issue but I don't know if mine could be considered normal behaviour: even if net.inet.tcp.reass.curesegments is set to 1680 and does not icrease, the output of vmstat -z shows a high tcpreass fail value that I don't remember in previous (8-STABLE) builds: ITEM SIZE LIMIT USED FREE REQ FAIL SLEEP [...] socket: 680, 25602, 740, 754, 4345747, 0, 0 unpcb: 240, 25600, 85, 459, 148595, 0, 0 ipq: 56, 819, 0, 378, 16126, 0, 0 udp_inpcb: 392, 25600, 24, 376, 222036, 0, 0 udpcb: 16, 25704, 24, 480, 222036, 0, 0 tcp_inpcb: 392, 25600, 719, 2391, 3958901, 0, 0 tcpcb: 976, 25600, 625, 707, 3958901, 0, 0 tcptw: 72, 5150, 93, 2357, 1486035, 0, 0 syncache: 152, 15375, 2, 398, 1587985, 0, 0 hostcache: 136, 15372, 1490, 4922, 119374, 0, 0 tcpreass: 40, 1680, 1680, 0, 108934,4800302, 0 sackhole: 32, 0, 0, 404, 134750, 0, 0 [...] System is FreeBSD 9.0-PRERELEASE #8 r227705 Bye, On Thu, Nov 24, 2011 at 8:02 AM, Kris Bauer <kristoph.bauer@gmail.com> wrote:> Hello, > > I am currently experiencing an issue with FreeBSD 9.0-RC2 r227852 where the > net.inet.tcp.reass.curesegments value is constantly increasing (and not > descreasing when there is nominal traffic with the box). ?It is causing tcp > slowdowns as described with kern/155407: > > Exhausted net.inet.tcp.reass.maxsegments block recovering tcp session (for > this socket and any other socket waiting for retransmited packets). After > exhausted net.inet.tcp.reass.maxsegments allocation new entry in tcp_reass > failed (for this socket and any other socket waiting for retransmited > packets). > > I have increased the reass.maxsegments value to 16384 to temporarily avoid > the problem, but the cursegments number keeps rising and it seems it will > occur again. > > Is this an issue that anyone else has seen? ?I can provide more information > if need be. > > Thanks, > Kris > _______________________________________________ > freebsd-stable@freebsd.org mailing list > http://lists.freebsd.org/mailman/listinfo/freebsd-stable > To unsubscribe, send any mail to "freebsd-stable-unsubscribe@freebsd.org" >
On 24.11.2011. 8:02, Kris Bauer wrote:> Hello, > > I am currently experiencing an issue with FreeBSD 9.0-RC2 r227852 where the > net.inet.tcp.reass.curesegments value is constantly increasing (and not > descreasing when there is nominal traffic with the box). It is causing tcp > slowdowns as described with kern/155407: > > Exhausted net.inet.tcp.reass.maxsegments block recovering tcp session (for > this socket and any other socket waiting for retransmited packets). After > exhausted net.inet.tcp.reass.maxsegments allocation new entry in tcp_reass > failed (for this socket and any other socket waiting for retransmited > packets). > > I have increased the reass.maxsegments value to 16384 to temporarily avoid > the problem, but the cursegments number keeps rising and it seems it will > occur again. > > Is this an issue that anyone else has seen? I can provide more information > if need be.Is your configuration different than the default in some way? Do you use a firewall? Multithreaded netisr? One of the new TCP congestion control modules?
On Thu, Nov 24, 2011 at 4:06 PM, Stefan Bethke <stb@lassitu.de> wrote:> Am 24.11.2011 um 21:30 schrieb Kris Bauer: > > > On Thu, Nov 24, 2011 at 1:20 PM, Raul <raul@turing.b2n.org> wrote: > > > > I am seeing the same sorts of things in netstat & vmstat: > > > > # netstat -s -p tcp |grep mem > > 742935 discarded due to memory problems > > > > # vmstat -z |grep tcpreass > > tcpreass: 40, 16464, 16340, 124, 131485,955443, 0 > > Same here: > root@diesel:~# netstat -s -p tcp |grep mem > 529211 discarded due to memory problems > root@diesel:~# vmstat -z |grep tcpreass > tcpreass: 40, 1680, 1679, 1, 118846,831450, 0 > root@diesel:~# uname -a > FreeBSD diesel.lassitu.de 9.0-PRERELEASE FreeBSD 9.0-PRERELEASE #20: Fri > Nov 18 21:57:59 CET 2011 root@diesel.lassitu.de:/usr/obj/usr/src/sys/DIESEL > amd64 > root@diesel:~# uptime > 11:01PM up 5 days, 23:15, 1 user, load averages: 0.14, 0.04, 0.01 > root@diesel:~# svn info /usr/src > Path: /usr/src > Working Copy Root Path: /usr/src > URL: http://mirror.hanse.de/svn/freebsd/base/stable/9 > Repository Root: http://mirror.hanse.de/svn/freebsd/base > Repository UUID: ccf9f872-aa2e-dd11-9fc8-001c23d0bc1f > Revision: 227665 > Node Kind: directory > Schedule: normal > Last Changed Author: fabient > Last Changed Rev: 227664 > Last Changed Date: 2011-11-18 15:41:48 +0100 (Fri, 18 Nov 2011) > > I regularly copy large files off my Tivo trans-atlantic (125ms RTT), and > TCP connections currently stall after about 500 megs, never recovering. I > suspect this is connected, as it started immediately after upgrading the > machine to 9-stable. > > As far as I can tell, the problem does not exist with 8-stable. > > > Stefan > > -- > Stefan Bethke <stb@lassitu.de> Fon +49 151 14070811 > > > > _______________________________________________ > freebsd-stable@freebsd.org mailing list > http://lists.freebsd.org/mailman/listinfo/freebsd-stable > To unsubscribe, send any mail to "freebsd-stable-unsubscribe@freebsd.org" >100-150ms RTT trans-atlantic transfers is what I saw (largely) driving up the reass.cursegments value.
Have you tried disabling the tcp offload features of your NIC? Adrian
On 11/24/11 18:02, Kris Bauer wrote:> Hello, > > I am currently experiencing an issue with FreeBSD 9.0-RC2 r227852 where the > net.inet.tcp.reass.curesegments value is constantly increasing (and not > descreasing when there is nominal traffic with the box). It is causing tcp > slowdowns as described with kern/155407: > > Exhausted net.inet.tcp.reass.maxsegments block recovering tcp session (for > this socket and any other socket waiting for retransmited packets). After > exhausted net.inet.tcp.reass.maxsegments allocation new entry in tcp_reass > failed (for this socket and any other socket waiting for retransmited > packets). > > I have increased the reass.maxsegments value to 16384 to temporarily avoid > the problem, but the cursegments number keeps rising and it seems it will > occur again. > > Is this an issue that anyone else has seen? I can provide more information > if need be.Thanks Kris, Raul and Stefan for the reports, I'll look into this. Cheers, Lawrence
El 25/11/2011 0:35, Adrian Chadd escribi?:> Have you tried disabling the tcp offload features of your NIC?In my case, there is no tcp on the ethernet interface. It is pppoe (mpd / netgraph) so no fancy hardware acceleration there. [...] %ifconfig ng0 | head -n1 ng0: flags=88d1<UP,POINTOPOINT,RUNNING,NOARP,SIMPLEX,MULTICAST> metric 0 mtu 1492 [...] Regards, Ra?l.