Hi Gregor! On Sat, 14 Apr 2012 22:03:59 +0200, gregor herrmann wrote: > On Fri, 13 Apr 2012 16:49:19 +0200, Luca Capello wrote: >> First, something which is related, but not critical: IMHO iodined should >> be started *after* any DNS server, simply because in most cases the >> topdomain argument must be resolved via DNS. Attached a Git patch >> against the Debian repository: [...] > Here I have a question: It obviously makes sense to start iodined > after a named, but what happens if the machine doesn't run a > nameserver (which happens probably more often for iodined)? [...] > Or not. Trying it and looking throught the generated > /etc/init.d/.depend.* files looks good (and also other initscripts > have named in Required-Start and work although I don't have something > providing $named here.
Same here on my laptop, the biggest example being Postfix: ===== # grep named /etc/init.d/* /etc/init.d/cron:# Should-Start: $network $named slapd autofs ypbind nscd nslcd /etc/init.d/cron:# Should-Stop: $network $named slapd autofs ypbind nscd nslcd /etc/init.d/fetchmail:# Should-Start: $mail-transport-agent exim4 $named /etc/init.d/nslcd:# Should-Start: $named slapd /etc/init.d/postfix:# Required-Start: $local_fs $remote_fs $syslog $named $network $time /etc/init.d/postfix:# Required-Stop: $local_fs $remote_fs $syslog $named $network /etc/init.d/rsync:# Should-Start: $named /etc/init.d/screen-cleanup:# Script to remove stale screen named pipes on bootup. /etc/init.d/xdm:# Should-Start: xfs $named slapd hal /etc/init.d/xdm:# Should-Stop: xfs $named slapd hal # ===== >> Second, as suggested by Martin in #433528, /etc/default/iodine should be >> moved to /etc/default/iodined (which would also reflect the name of the >> init.d script). > > Yes, that's all a bit inconsistent ... I will come back if I find some time for it. >> Finally, the real problem. > >> While checking the upstream Git repository for relevant changes, I >> discovered that in configurations similar to mine, i.e. iodined and >> dnsmasq on different interfaces, there is no need to run the two daemons >> on separate ports: >> >> commit a62ae8e562d0e9e5729bb4f469bc900074930efc >> Author: Erik Ekman <[email protected]> >> Date: Thu Jul 16 08:44:04 2009 +0000 >> Subject: Allow bind port and listen port to be the same if listenip >> doesnt include localhost. Also remove newlines from warnx > > Great finding! This was my opinion as well, which however does not solve the problem I found for setups which needs both daemon running on the same interface. >> Actually, there is no mention of this in the iodine manpage, so attached >> a patch against the upstream Git repository (no mention in README to >> avoid naive usage): [...] >> Indeed, with this feature and the following configuration iodined does >> not segfault anymore: > > Very cool. > Have you forwarded this patch already or shall I do it? I have not forwarded it, so feel free to do it :-) Thx, bye, Gismo / Luca
pgpnye7AxQFzx.pgp
Description: PGP signature
