* Peer Stefan wrote on Mon, Jan 20, 2003 at 13:07 +0100:
The rest is done by the DNS and its "round robin" should give you a simple kind of load balancing, if both systems are up.
And makes 50% of the requests/client fail if one of two systems is down. Load balancing, but not failsafe. I don't think that it's a good idea to try to dynamic update DNS since this wouldn't help much due to caching. I think, this could be better solved with some alias IPs - in case of a crash, the other server get's it's IP. In practice, sometimes the crashed server still replies ARP, so take care :)
Be carefull with DNS doing the round robin - not every dns resolver can handle more than one ip-address for one name. E.g. Windows 9x/ME strips off all additional ip-addresses. It only uses the first ip-address it gets.
Yes, and that's no problem since DNS servers should do roundrobbing by sending any of the RRs first. So the same client usually makes all requests to the same servers, but after all it's still pretty balanced even with 24 bit windows :)
Then there is such a thing called name cache. All resolved hosts are stored within this cache for performance reasons. If such a "bad" net-member [...] "timeout [...] "
Yes, that's why it isn't possible to change DNS entries in failure cases. oki, Steffen -- Dieses Schreiben wurde maschinell erstellt, es trägt daher weder Unterschrift noch Siegel.