id	summary	reporter	owner	description	type	status	priority	milestone	component	version	severity	resolution	keywords	cc	mlocked
10288	FE Won't Connect to BE in 0.25pre if bindv6only = 1	Bill Meek <keemllib@…>	beirdo	"If a user upgrades to 0.25{pre}, and their distribution sets bindv6only = 1,
their frontends (combined or remote) will not connect to their backend.

1st, thanks to Ian Clark (mrrooster) and Erich (mepisrocks) for their help
in diagnosing this. Their Debian and Mepis systems (respectively) were
failing and had bindb6only = 1. 

Users can check their system by typing:
{{{
sysctl -a 2>/dev/null | grep bindv6only
}}}
To duplicate:
{{{
On a master backend:
Run 0.25pre
Verify that IPv4 addresses are set for the ...ServerIPs
Add /etc/sysctl.d/mythtv-bind-test.conf
Put: net.ipv6.bindv6only = 1 in that file
Reboot
Start a mythfrontend running
}}}

The FE will get the ""Could not connect to the master backend server.""
pop up.

Solution 1 (probably correct, easily overlooked):

Users / Distributions should not set bindv6only = 1. Note: #9820, #9840 and #10165
as well as posts on the -users list say this is already a problem.

Most likely solved by searching for bindv6only in files under /etc/sysctl.d and
removing it, commenting it out or setting it to 0.

Solution 2 (delay the problem):

Tested under v0.25pre-4146-g633370e.

The attached patch will prevent binding to IPv6 '''as long as''' the user
hasn't configured any IPv6 routable addresses on their system.  It does
this by ignoring Host (::1) and Link Local (fe80::) addresses that are
automatically configured by the system.

If/when a user adds a Global IPv6 address, the FE will not connect if
bindv6only = 1.

Solution 3 (bindv6only check/error message)

Included in the patch (in the #ifdef at the end) is a test of the
bindv6only setting. I used the /proc/sys interface only because in
the ubuntu man pages it says sysctl(2) may go away.
"	Patch - Bug Fix	closed	minor	unknown	MythTV - General	Master Head	medium	Won't Fix			0