On Nov 21, 2012, at 11:40 , intrigeri <intrig...@debian.org> wrote: > Kenyon Ralph wrote (23 Jun 2012 20:17:28 GMT) : >>> When you specify a target via hostname which has both a v6 and v4 >>> address munin tries to connect to the v6 address first and if this >>> fails bails out completely. This skips data collection for this >>> host, resulting in lost data points. > > Net::Server changed quite a lot in Debian since you first reported > this bug; e.g. it now ships with the new upstream IPv6 support, > instead of our old buggy Debian-specific IPv6 patch. > > So, can you please try to reproduce this issue on current Debian > testing/sid?
Just verified with munin 2.0.6-1 on wheezy, still broken: http://nopaste.narf.at/show/TANDm4LYv2byzg53eFfg/ The error actually is not in the Net::Server package but in the use of IO::Socket::INET6. /usr/share/perl5/Munin/Master/Node.pm:84 if ($uri->scheme eq "munin") { $self->{reader} = $self->{writer} = IO::Socket::INET6->new( PeerAddr => $uri->host, PeerPort => $self->{port} || 4949, LocalAddr => $config->{local_address}, Proto => 'tcp', Timeout => $config->{timeout} The lack of the "MulitHomed" parameter causes IO::Socket::INET6 to give preferential treatment to AAAA-addresses and fail if they don't yield a socket - when enabling MultiHomed IO::Socket::INET6 will try all returned addresses. See http://search.cpan.org/~shlomif/IO-Socket-INET6-2.69/lib/IO/Socket/INET6.pm#CONSTRUCTOR for detailed information. Just hacked MultiHomed in myself, this works fine when having dual stack munin-nodes only listening on v4 addresses. all the best, Michael
