Date: Fri, 22 Apr 2005 02:22:04 -0400 From: "Michael W. Lucas" <mwlucas@blackhelicopters.org> To: "Michael W. Lucas" <mwlucas@blackhelicopters.org> Cc: mlaier@freebsd.org Subject: Re: CARP-related error on 5.4RC3 Message-ID: <20050422062204.GA525@bewilderbeast.blackhelicopters.org> In-Reply-To: <20050422053731.GA99242@bewilderbeast.blackhelicopters.org> References: <20050422053731.GA99242@bewilderbeast.blackhelicopters.org>
next in thread | previous in thread | raw e-mail | index | archive | help
(copying mlaier@ because FreeBSD carp seems to be his baby, and we're trying to squeeze 5.3 out) Sorry to follow up on my own post, but more CARP weirdness on 5.4RC3. I have two traffic shapers, both i386 with fresh 5.4RC3 installs, one master and one backup. When I reboot the master, the backup quickly takes over -- just as you would expect. When the master finishes booting, however, the backup doesn't release one of the CARP interfaces. Even if I wait several minutes, the backup looks like this: pfsync0: flags=41<UP,RUNNING> mtu 1348 pfsync: syncif: fxp0 maxupd: 128 carp0: flags=41<UP,RUNNING> mtu 1500 inet xxx.yy.zzz.20 netmask 0xffffffc0 carp: MASTER vhid 1 advbase 1 advskew 100 carp1: flags=41<UP,RUNNING> mtu 1500 inet xxx.yy.zzz.129 netmask 0xffffff80 carp: BACKUP vhid 2 advbase 1 advskew 100 Note that carp0 is a master, while carp1 is a slave. Meanwhile, over on the master we have: pfsync0: flags=41<UP,RUNNING> mtu 1348 pfsync: syncif: fxp0 maxupd: 128 carp0: flags=41<UP,RUNNING> mtu 1500 inet 198.88.118.20 netmask 0xffffffc0 carp: MASTER vhid 1 advbase 1 advskew 0 carp1: flags=41<UP,RUNNING> mtu 1500 inet 198.88.118.129 netmask 0xffffff80 carp: MASTER vhid 2 advbase 1 advskew 0 So, two carp0 masters. This is not good. Both machines have the same sysctl settings: net.inet.carp.allow: 1 net.inet.carp.preempt: 1 net.inet.carp.log: 1 net.inet.carp.arpbalance: 0 ==ml On Fri, Apr 22, 2005 at 01:37:31AM -0400, Michael W. Lucas wrote: > > On a new backup traffic shaper I have the following routing table: > > Internet: > Destination Gateway Flags Refs Use Netif Expire > default xxx.yy.zzz.1 UGS 0 13708 carp0 > 127.0.0.1 127.0.0.1 UH 0 0 lo0 > ... > > Note the "netif" for my default route. When carp0 is in a backup > state, I cannot reach my default route from this machine! As you > might guess, this makes life difficult. > > On the master machine, though, the default route points at a real > interface and I can reach the Internet: > > Internet: > Destination Gateway Flags Refs Use Netif Expire > default xxx.yy.zzz.1 UGS 0 4563 em0 > 127.0.0.1 127.0.0.1 UH 0 0 lo0 > ... > > The machines have very different hardware: the master is a new Dell > 1650 with em cards, the backup is a desktop machine with xl cards. > The software is identical, however. > > In the interface list for the master machine, the em cards appear > above the carp interfaces. On the backup machine, the carp interfaces > appear before the xl interfaces. Might the default route simply be > attaching to the first appropriate interface it finds? > > > Since I have to get this machine in production tonight, I'm going to > try to hack around this, maybe with a shell script that creates and > ifconfigs the carp interfaces after the real interfaces are > configured. It's definitely odd, though, and figured it should be > reported. > > > > > > > -- > Michael W. Lucas mwlucas@FreeBSD.org, mwlucas@BlackHelicopters.org > http://www.BlackHelicopters.org/~mwlucas/ > Latest book: Cisco Routers for the Desperate > http://www.CiscoRoutersForTheDesperate.com -- Michael W. Lucas mwlucas@FreeBSD.org, mwlucas@BlackHelicopters.org http://www.BlackHelicopters.org/~mwlucas/ Latest book: Cisco Routers for the Desperate http://www.CiscoRoutersForTheDesperate.com
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20050422062204.GA525>