Date: Tue, 6 Feb 2001 09:42:00 -0800 (PST) From: Matt Dillon <dillon@earth.backplane.com> To: Rik van Riel <riel@conectiva.com.br> Cc: Dan Phoenix <dphoenix@bravenet.com>, Alfred Perlstein <bright@wintelcom.net>, Jos Backus <josb@cncdsl.com>, freebsd-hackers@FreeBSD.ORG Subject: Re: qmail IO problems Message-ID: <200102061742.f16Hg0x62032@earth.backplane.com> References: <Pine.LNX.4.21.0102061112060.1535-100000@duckman.distro.conectiva>
next in thread | previous in thread | raw e-mail | index | archive | help
: :On Mon, 5 Feb 2001, Matt Dillon wrote: : :> :file: table is full : :> :looutput: mbuf allocation failed :> :nfs server 172.16.0.101:/bravenet1/home: is alive again : :> This sheds a considerable amount of light on the problems... :> methinks you may have a low 'maxusers' setting in the kernel :> config. Read on. : :Linux had problems with errors like this too, with :kernel 2.0 and 2.2 when used under heavy load. : :In kernel 2.4 this has been solved by simply having :the kernel allocate (and free) these structures on :demand ... would that be an idea for FreeBSD ? : :Rik :-- :Linux MM bugzilla: http://linux-mm.org/bugzilla.shtml : :Virtual memory is like a game you can't win; :However, without VM there's truly nothing to lose... : : http://www.surriel.com/ :http://www.conectiva.com/ http://distro.conectiva.com/ Yes, it's very doable. There are only a few subsystems which actually scale based on 'maxusers'. File descriptors, sendfile buffers, and network mbuf clusters. I think a good temporary fix would be to change the absurdly small default maxusers of 32 to something more reasonable, like 128. -Matt To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200102061742.f16Hg0x62032>