Bug#954321: in the Debian AWS-images cloud-init network fails if ipv6 is disabled

2020-03-20 Thread Bastian Blank
Control: reassign -1 debian-cloud-images
Control: severity -1 minor
Control: tags -1 ipv6,help

Hi Bas

On Fri, Mar 20, 2020 at 09:00:08AM +0100, Bas Zoetekouw wrote:
> On these machines, ipv6 is disabled by setting
> /proc/sys/net/ipv6/conf/all/disable_ipv6 to 1. This causes the network
> to fail:

No, our AWS images don't use disable_ipv6.  If you modify this setting,
you of course need to adopt the network config to cope with it.

Please provide a patch if you think we can do better.

Bastian

-- 
Those who hate and fight must stop themselves -- otherwise it is not stopped.
-- Spock, "Day of the Dove", stardate unknown



Bug#954321: in the Debian AWS-images cloud-init network fails if ipv6 is disabled

2020-03-20 Thread Bas Zoetekouw
package: cloud.debian.org

(I previously reported this bug again cloud-init, see
https://bugs.launchpad.net/cloud-init/+bug/1863773, but it seems it is
instead an issue with the Debian images themselves).

I'm using cloud-init on a Debian machien in AWS using the official
Debian buster images
(https://wiki.debian.org/Cloud/AmazonEC2Image/Buster
)

On these machines, ipv6 is disabled by setting
/proc/sys/net/ipv6/conf/all/disable_ipv6 to 1. This causes the network
to fail:

bas@machine:/etc/network$ sudo ifup ens5
Internet Systems Consortium DHCP Client 4.4.1
Copyright 2004-2018 Internet Systems Consortium.
All rights reserved.
For info, please visit https://www.isc.org/software/dhcp/


Listening on LPF/ens5/02:45:ca:0d:93:76
Sending on LPF/ens5/02:45:ca:0d:93:76
Sending on Socket/fallback
DHCPREQUEST for 10.0.16.20 on ens5 to 255.255.255.255 port 67
DHCPACK of 10.0.16.20 from 10.0.16.1
RTNETLINK answers: File exists
bound to 10.0.16.20 -- renewal in 1796 seconds.

Could not get a link-local address
run-parts: /etc/network/if-pre-up.d/cloud_inet6 exited with return code 1
ifup: failed to bring up ens5

Even worse, on boot the machines do get a proper ipv4 address, but
(apparently because setting up the network fails), the dhcp (v4) daemon
is stopped, and the machines "mysteriously" drop off the net after the
lease ends.