If all the DNS servers really adhere to the RFCs and really do cache using the TTLs stated by the zone's SOA server, this might be less of a headache than it is in reality. Unfortunately not every ISP is using BIND8, and some are using much worse than DNS server implementations that the wintel services out there which completely disregard TTL's in the zone records and make up there own mind on how long they'll cache. Besides if you are going to go through the trouble and cost oof building out web services in multiple locations, any downtime to any enduser due to the architecture is unacceptible. Think about it... Imagine .... You (yourself) are shelling out somewhere in the neighborhood of a grand a month for shelf space, then throw on bandwidth, all the programming and configuration investment (500 - $10,000 [depending on your sucker level]) for the set up... And you gone thru all this trouble... As a businessman, are you *really* comfortable with an occasional 30-60 minute down time for RFC compliant DNS caches and 24-72 hour for non-compliant DNS caching? I've seen the happiness levels of quite of few big sites over this type of approach go to hell. It's cheap going in, but they end up digging in their pockets for that ole F5 solution afterwards. Don't get me wrong, I think F5 is WAY overpriced and as soon as GNU gets hold of it and maybe Vixie pushes some of it into BIND we can all sleep without our pagers with a little less trepidation. -Karyn -----Original Message----- From: Richard A. Steenbergen [mailto:ras@e-gerbil.net] Sent: Wednesday, July 05, 2000 10:39 AM To: Karyn Ulriksen Cc: 'nanog@merit.edu' Subject: RE: bad idea? On Wed, 5 Jul 2000, Karyn Ulriksen wrote:
Riigghhhttttt..... But if the server goes down and the IP is unreachable, DNS will never know and all his structure goes for naught...
Thats actually a good thing, at least for the short-term. If the IP is down hard and isn't coming back up (for example, the server crashed), thats a different story. A short TTL works, if you don't mind it offsetting the benefits of DNS caching. Its an all around grody hack, but its still better then other alternatives. It also helps if the DNS "server" for this load balanced sub-domain is integrated with the box, for example a Foundry ServerIron which can perform health checks, and withdraw the announcement if its no longer viable to get answers from this server. As long as the grody-ness stays restricted to this confined area, and doesn't break anything else... I can live with it. -- Richard A Steenbergen <ras@e-gerbil.net> http://www.e-gerbil.net/humble PGP Key ID: 0x138EA177 (67 29 D7 BC E8 18 3E DA B2 46 B3 D8 14 36 FE B6)