Re: [Ganglia-developers] 2.6.0 test

[Brooks D. <br...@on...>]

On Thu, Jul 15, 2004 at 12:10:02PM -0700, Matt Massie wrote:
> On Thu, 2004-07-15 at 10:29, Brooks Davis wrote:
>=20
> > Partial success on FreeBSD 5-CURRENT, unless this is supposed to be a v6
> > only test.  It's only binding to :: not 127.0.0.1.  See below for my te=
sts.
> >=20
> > Some thoughts on using apr.  In general, I like the idea.  The
> > config.layout feature looks really nice!  However, apr is big so it
> > would be nice it we could have an option to use an external verion of
> > apr, for instance when built in the FreeBSD ports collection.  The
> > tarball is also rather gigantic with apr.  Perhaps two distrubtions
> > would be in order, one with all the deps for lazy people/broken package
> > systems, and one with just the core for people who already have the deps
> > are can get them easily.
>=20
> that is already in the design of this framework.  this is a request that
> preston smith (the other freebsd guru) asked for quiet a while back and
> it makes a lot of sense.
>=20
> % ./configure --help
> ...
>    --with-apr=3DDIR|FILE     prefix for installed APR, path to APR build
> tree,
>                           or the full path to apr-config
>   --with-apr-util=3DDIR     prefix for installed APU, or path to APU build
> tree
>=20
> this allows package maintainers to decide if they want to use the
> packaged apr/apr-util or instead create a dependency in their package on
> the external library.

Excellent!

> if you are worried about the size of the package... i guess we could
> have a (with srclib and without srclib setup).  i don't think a couple
> meg download should be much of a problem though... even 5 megs.

I agree and disagree.  On one hand,it's only a few MB out of .5+GB of
dist files on my machine.  On the other hand, that's a few more minutes
I can't use very productively every time I have to download the source
(I'm stuck with IDSL here).

> > [10:22am] brooks@minya (~/working/ganglia/ganglia-2.6.0/gmond): socksta=
t| grep gmon
> > brooks   gmond      23960 3  tcp6   *:8021                *:*
> > [10:22am] brooks@minya (~/working/ganglia/ganglia-2.6.0/gmond): telnet =
localhost 8021
> > Trying ::1...
> > Connected to localhost.
> > Escape character is '^]'.
> >         Server socket: ::1:8021 -> ::1:49506
> > Connection closed by foreign host.
>=20
> it looks like localhost is resolving to an IPv6 address.
>=20
> > [10:22am] brooks@minya (~/working/ganglia/ganglia-2.6.0/gmond): telnet =
:: 8021  Trying ::...
> > Connected to ::.
> > Escape character is '^]'.
> >         Server socket: ::1:8021 -> ::1:49507
> > Connection closed by foreign host.
>=20
> good.
>=20
> > [10:22am] brooks@minya (~/working/ganglia/ganglia-2.6.0/gmond): telnet =
0  8021
> > Trying 0.0.0.0...
> > telnet: connect to address 0.0.0.0: Connection refused
> > telnet: Unable to connect to remote host
>=20
> expected.
>=20
> > [10:23am] brooks@minya (~/working/ganglia/ganglia-2.6.0/gmond): telnet =
127.0.0.1 8021
> > Trying 127.0.0.1...
> > telnet: connect to address 127.0.0.1: Connection refused
> > telnet: Unable to connect to remote host
>=20
> this is strange and not what i would expect (or found on other operating
> systems).  hmmmm.  are you running any ip filters on the box? =20

I found the problem.  FreeBSD is intentionally violating RFC2553 because
they (KAME) feel it is dangerous.  There are two solutions to this.
One is to enable IPv4-mapped IPv6 address support via a socket option.
The other is to explicitly try to bind both v4 and v6 sockets.

This is documented in the "Interaction between IPv4/v6 sockets" section
of the inet6(4) manpage:

http://www.freebsd.org/cgi/man.cgi?query=3Dinet6

The sockopt is IPV6_V6ONLY which is documented in ip6(4):

http://www.freebsd.org/cgi/man.cgi?query=3Dip6

-- Brooks

--=20
Any statement of the form "X is the one, true Y" is FALSE.
PGP fingerprint 655D 519C 26A7 82E7 2529  9BF0 5D8E 8BE9 F238 1AD4