From mboxrd@z Thu Jan 1 00:00:00 1970 From: "Adrian C." Subject: Re: SSL Certificate signing problem Date: Sun, 10 Oct 2004 03:24:10 +0300 Sender: linux-admin-owner@vger.kernel.org Message-ID: <60a7468904100917247ae77c18@mail.gmail.com> References: <60a74689041009164250458ee2@mail.gmail.com> <4168804A.9020200@kjchome.homeip.net> Reply-To: "Adrian C." Mime-Version: 1.0 Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <4168804A.9020200@kjchome.homeip.net> List-Id: Content-Type: text/plain; charset="us-ascii" To: "Kevin J. Cummings" Cc: linux-admin If i assign metric 1 to 1st gway and 2 to 2nd it never falls back from gateway with best metric. Even if it goes down it still sticks to it. Something is terribly wrong. I am running Slackware 10. On Sat, 09 Oct 2004 20:20:26 -0400, Kevin J. Cummings wrote: > > > Adrian C. wrote: > > Hello. I'm trying to setup a simple failover between 2 gateways on kernel 2.6.2 > > Here it goes. > > just one interface for everything: eth0 > > route add default gw 192.168.1.1 > > route add default gw 192.168.2.1 > > > > let's say i ping gmail.com and i kill the 192.168.1.1 machine. ping > > stops for about 2 minutes then the next gateway is used and the ping > > comes back to live. The only problem here is that it forgets to NAT my > > clients via the new gway. At least that's my only explanation why it > > stops NATting. Masquerading is done without a -d so destination is > > any. What can be done here? > > Also please let me know the files in which i should modify fallback > > timeout for routes. I need a route check every 10 seconds or so. > > One more thing, if 192.168.1.1 comes back to live i would like to > > become the preferred gateway no matter if 192.168.1.2 is alive and > > used by kernel. Is this solved by metrics? > > I thought that this is what Metrics were supposed to do. If you > *prefer* the .1.1 route, assign it a better Metric than the .2.1 route. > Then, when the .1.1 route goes down, the packets should be immediately > re-routed to the .2.1 interface, and when the .1.1 comes back up, the > first route should work again. THat's *my* understanding of how Metrics > are supposed to work. > > I'm not an expert, and I'm not sure about what happens after a TCP > connection is already established and then an interface fails (or > restores) whether *that* connection will continue to use the previous > routing or not. > > -- > Kevin J. Cummings > kjchome@rcn.com > cummings@kjchome.homeip.net > cummings@kjc386.framingham.ma.us >