Date: Mon, 24 Jan 2005 16:41:36 +0300 From: Evgeny Dolgopiat <dolgop@mccinet.ru> To: "Pawel Jakub Dawidek" <pjd@FreeBSD.org> Cc: freebsd-net@FreeBSD.ORG Subject: Re: New failure detection algorithm for ng_one2many(4). Message-ID: <200501241641.36863.dolgop@mccinet.ru>
next in thread | raw e-mail | index | archive | help
> Hello. > > Patch below adds new failure detection algorithm for ng_one2many(4). > For now, the only such algorithm doesn't detect failures, one have > to show active links manually. > > http://garage.freebsd.pl/patches/netgraph_one2many.patch > > The way how to use this module (ng_one2many) is described in manual > page, to made use of new algorithm one should execute: > > ngctl msg fxp0:upper setconfig "{ xmitAlg=3D1 failAlg=3D2 >interval=3D5 }" > > Value '5' after interval means: check interfaces every 5 seconds. >If interface will be down it will not be used. > >Note that 'one' interface have to be always avaliable. > >Any comments, etc. are of course welcome Hello. Did you see my patch? http://docs.freebsd.org/cgi/getmsg.cgi?fetch=92946+0+archive/2005/freebsd-net/20050116.freebsd-net Heartbeat signal has some advantages as compared to ioctl function. All subnet containing failed element (card or hub) marked as failed. Therefore packets sended from other hosts to host, holding failed element, wouldn't lost. Other advantage: now heartbeat algorithm working independently of the layer on which one2many operates (not only for ethernet). Some time ago (october 2003) I sent first version of this patch in mailing list and got some replies from people, using STABLE (4.X). But this patch could be used only for 5.X, not stable at that time. Now 5.3 is stable and this patch can be used for it.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200501241641.36863.dolgop>