CentOS - Sep 2015 - (?) Mailman VERY slow with IPv6 (with work-around)

I recently stood up an EL7 box with Mailman for a few lists I run for
some friends.  My old install, on an EL6 system, ran with no issues for
several years but I was induced to upgrade by a "hardware casualty" on
the old system.  I was going to have to rebuild anyway, so why not take
it as an opportunity to try EL7?

The build went fine and I was able to migrate the lists over with no
issues, but once I got there just about everything to do with Mailman
operations were painfully slow.  For example, "list_lists" took 5
seconds of "real" time.  I was used to it taking _much_ less as I only
have about 6 lists.  This affected both the command-line Mailman tools
and the web interface.  My first inclination was to blame Python, but
other code executed just fine with it.

While testing I tried an strace of list_lists and found that it was
timing out on a read operation to a socket to the Avahi daemon
(/var/run/avahi-daemon/socket) while trying to resolve the link-local
IPv6 address.  Having flashbacks to Sendmail stalling on DNS issues I
decided to try fixing resolution first.

As a test I put the link-local address into my /etc/hosts file with a
localized name.  Running list_lists then took about 0.19 seconds
"real"
time!  The web interface also changed from painfully slow to it's
previous behaviour on EL6.

I imagine just turning off IPv6 would work as well, but I have an actual
use case that is a lot easier with it turned on.  I don't know if
anybody else has seen this, but thought it might be handy for someone else.
-- 
Jay Leafey - Memphis, TN
jay.leafey at mindless.com

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 17/09/15 04:16, Jay Leafey wrote:
> I recently stood up an EL7 box with Mailman for a few lists I run
> for some friends.  My old install, on an EL6 system, ran with no
> issues for several years but I was induced to upgrade by a
> "hardware casualty" on the old system.  I was going to have to
> rebuild anyway, so why not take it as an opportunity to try EL7?
> 
> The build went fine and I was able to migrate the lists over with
> no issues, but once I got there just about everything to do with
> Mailman operations were painfully slow.  For example,
"list_lists"
> took 5 seconds of "real" time.  I was used to it taking _much_
less
> as I only have about 6 lists.  This affected both the command-line
> Mailman tools and the web interface.  My first inclination was to
> blame Python, but other code executed just fine with it.
> 
> While testing I tried an strace of list_lists and found that it
> was timing out on a read operation to a socket to the Avahi daemon 
> (/var/run/avahi-daemon/socket) while trying to resolve the
> link-local IPv6 address.  Having flashbacks to Sendmail stalling on
> DNS issues I decided to try fixing resolution first.
> 
> As a test I put the link-local address into my /etc/hosts file with
> a localized name.  Running list_lists then took about 0.19 seconds
> "real" time!  The web interface also changed from painfully slow
to
> it's previous behaviour on EL6.
> 
> I imagine just turning off IPv6 would work as well, but I have an
> actual use case that is a lot easier with it turned on.  I don't
> know if anybody else has seen this, but thought it might be handy
> for someone else.
> 
Thanks a lot for having shared this. I haven't tried (yet) to test
mailman on CentOS 7 and the only one we have is on CentOS 6, and
(unfortunately) no ipv6 addr, but I do remember having had to do this
for ipv4 addr too (in /etc/hosts) to bypass the high number of dns
requests (in the past)

- -- 
Fabian Arrotin
The CentOS Project | http://www.centos.org
gpg key: 56BEC54E | twitter: @arrfab
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.22 (GNU/Linux)

iEYEARECAAYFAlX6ZHwACgkQnVkHo1a+xU4GyQCdFCkLZQNqV37iISS4Ejn1kqbg
xMAAoIrdK0Hmyx2uEk5vSSkWOdglH6xF
=+p+d
-----END PGP SIGNATURE-----

> While testing I tried an strace of list_lists and found that it
> >was timing out on a read operation to a socket to the Avahi daemon
> >(/var/run/avahi-daemon/socket) while trying to resolve the
> >link-local IPv6 address.
Is avahi even installed by default on CentOS 7?  I don't see it 
installed on my lab's desktop systems, let alone servers.