Date: Fri, 30 May 1997 08:59:53 -0700 (PDT) From: bigfoot@stomped.com To: freebsd-gnats-submit@FreeBSD.ORG Subject: kern/3719: cached routes flushed too quickly on busy systems Message-ID: <199705301559.IAA22223@hub.freebsd.org> Resent-Message-ID: <199705301600.JAA22272@hub.freebsd.org>
next in thread | raw e-mail | index | archive | help
>Number: 3719 >Category: kern >Synopsis: cached routes flushed too quickly on busy systems >Confidential: no >Severity: non-critical >Priority: low >Responsible: freebsd-bugs >State: open >Class: change-request >Submitter-Id: current-users >Arrival-Date: Fri May 30 09:00:02 PDT 1997 >Last-Modified: >Originator: Guy Gustavson >Organization: Reliant Net Services >Release: 2.1.5 >Environment: FreeBSD stomped.com 2.1.5-RELEASE FreeBSD 2.1.5-RELEASE #1: Wed May 28 19:02:54 CDT 1997 bigfoot@stomped.com:/usr/src/sys/compile/BFKERNEL i386 >Description: In module /usr/src/sys/netinet/in_rmx.c The maxium numbered of cached routes to allow before lowering the maxium alouded time to live on a route is hard coded at 128. This is probably to low for a heavily loaded system. >How-To-Repeat: Run a busy web server... like 50,000-100,000 page views a day. Sorry, but varied and heavy traffic is about the only way to get the expire messages. >Fix: On line 175 of /usr/src/sys/netinet/in_rmx.c Is have made this change /* int rtq_toomany = 128; /* 128 cached routes is ``too many'' int rtq_toomany = 512; /* 512 cached routes is ``too many'' according to me. GWG.*/ This is too high for a lightly loaded system with little memory, and the orginal is too low for a heavily loaded system. I suggest that the limit should probably be based on amount of real memory in the system, and somehow tied to the number of unique domains talked to each hour. >Audit-Trail: >Unformatted:
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199705301559.IAA22223>