
caudy at jhu
Nov 27, 2003, 8:13 AM
Post #2 of 4
(895 views)
Permalink
|
|
execute a command after getting/loosing a IP-Address?
[In reply to]
|
|
So, we don't provide any solutions to this problem within Wackamole.= =20 However, I'm sure someone else has had to deal with it, and might ans= wer=20 you with something they've already developed. (This is a big holiday= in=20 the US, which means a lot of people on the list probably won't be= =20 responding for a few days). The solution you proposed is probably the quickest and dirtiest way t= o=20 get what you want. If you wanted something more powerful, using the= =20 same mechanism to run a pair of shell-scripts "Acquire-script" and= =20 "Release-script" would give this to you. However, I'd be a little= =20 worried about doing so in a general situation, because of the privile= ges=20 with which Wackamole is run. I wouldn't recommend adding anything too complicated directly to=20 Wackamole, because of timing concerns. You might consider a=20 long-running daemon to handle this failover, that communicates=20 (something along the lines of "Acquire/Release address") with Wackamo= le=20 via named pipes or another Unix IPC mechanism. Hops this helps, Ryan Oliver F=E4nger wrote: > Hi list, >=20 > i've set up a round robin/failover scenario with squid. > Wackamole works fine, but my problem is that, in my case, > squid has to bind itself to a special address (not all 0.0.0.0). > So i have to send a kill -HUP to the squid process, after getting a= new > IP address, because squid does not listen to this IP. > The problem occurs with any other process, that binds to a special = IP=20 > and not > to all. >=20 > I've put a line in wackamole.c at the end of the Acquire function: > system ("kill -HUP `cat /var/run/squid.pid >/dev/null 2>&1` >/dev/n= ull=20 > 2>&1"); >=20 > But I don't like this very dirty solution, so i wanted to ask, > if there could be a better way to execute a command after > acquiring/releasing an ip address. >=20 > I can write the system stuff in c, but I would prefer a more=20 > configurable solution. >=20 >=20 > Thanks. >=20 > Kind regards, >=20 >=20 >=20 > Oliver F=E4nger >=20 >=20 >=20 >=20 >=20 >=20 > _______________________________________________ > wackamole-users mailing list > wackamole-users[at]lists.backhand.org > http://lists.backhand.org/mailman/listinfo/wackamole-users >=20 >=20 --=20 Ryan W. Caudy Center for Networking and Distributed Systems Department of Computer Science Johns Hopkins University
|