- From: Daniel O'Callaghan <danny@miriworld.its.unimelb.edu.au>
- Date: Fri, 30 Jun 1995 09:06:24 +1000 (EST)
- To: Kent Fitch <Kent.Fitch@its.csiro.au>
- Cc: Monica Berko <Monica.Berko@anu.edu.au>, www-proxy@www10.w3.org
On Thu, 29 Jun 1995, Kent Fitch wrote: > If an arp fails (indicating that at least the network stuff is > out-of-business), it reconfigures itself with the IP # of the main proxy > (copies a readied version of /etc/rc.d/rc.inet1 or equivalent with the > main proxy's IP #) and then reboots itself, coming up with the new #. > > Maybe a cheapo 486/pentium with a scsi running linux/bsd could provide a > "reasonable" service for a while at least, especially if the problem > occurs during the weekend/night when it is least likely to be rapidly fixed. This sounds pretty neat. Using an *86 machine with FreeBSD would allow you to ifconfig ed0 new-IP alias, to run two IP addresses on the one ethernet port. > > Using a machine that normally does other things as well is probably > precluded by the need to change the IP #. except with IP address aliases. > Things get more complicated if just the proxy httpd fails, but arps/pings > work, or if the main proxy magically recovers and reboots itself... It > would be nice if a really sick main proxy server could realize that it was > in trouble and just die cleanly. Mechanisms to do this could be > imagined, maybe relying on the backup proxy to maintain a TCP/IP > connection with the primary proxy on which it can send a logical signal > to "please shutdown now" when the need arises. Fun to do! I'm sure it could be worked out. Even if the sick main proxy server could renumber itself, too. But of course, the best answer is to coordinate with the guys who run DNS, and have DNS automatically reconfigured. Danny
Received on Thursday, 29 June 1995 19:06:45 UTC