Debian: Controller-0 failed to come up after unlock due to network config error

[Caio Bruchert]

Brief Description

Controller-0 failed to come up after unlock. Failed to configure mgmt and cluster on vlan interface.Both mgmt and cluster configured on same physical interface(pxwboot)

Severity
Major: System/Feature is usable

Steps to Reproduce

Fresh install (Debian)

Expected Behavior

Controller-0 to be up and available/online state

Actual Behavior

down, all services are down

Reproducibility

Seen once

System Configuration
Two node system, IPv6

Load info (eg: 2022-03-10_20-00-07)

Last Pass

First time for Debian Install

Timestamp/Logs

2022-10-03T14:13:34.667 Error: 2022-10-03 14:13:34 +0000 'ip addr replace abcd:205::1/64 dev vlan190 preferred_lft 0' returned 1 instead of one of [0]
2022-10-03T14:13:34.668 /usr/lib/ruby/vendor_ruby/puppet/util/errors.rb:157:in `fail'
2022-10-03T14:13:34.764 Error: 2022-10-03 14:13:34 +0000 /Stage[main]/Platform::Network::Addresses/Platform::Network::Network_address[cluster-host]/Exec[Configuring cluster-host IP address]/returns: change from 'notrun' to ['0'] failed: 'ip addr replace abcd:205::1/64 dev vlan190 preferred_lft 0' returned 1 instead of one of [0]
2022-10-03T14:13:34.775 Error: 2022-10-03 14:13:34 +0000 'ip addr replace abcd:204::1/64 dev vlan189 preferred_lft 0' returned 1 instead of one of [0]
2022-10-03T14:13:34.777 /usr/lib/ruby/vendor_ruby/puppet/util/errors.rb:157:in `fail'
2022-10-03T14:13:34.871 Error: 2022-10-03 14:13:34 +0000 /Stage[main]/Platform::Network::Addresses/Platform::Network::Network_address[mgmt]/Exec[Configuring mgmt IP address]/returns: change from 'notrun' to ['0'] failed: 'ip addr replace abcd:204::1/64 dev vlan189 preferred_lft 0' returned 1 instead of one of [0]

Alarms

N/A

Test Activity

Debian Install

Workaround

N/A

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.opendev.org/c/starlingx/config/+/861205

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.opendev.org/c/starlingx/config/+/861205
Committed: https://opendev.org/starlingx/config/commit/a13a27c767070770b17bf4f9eb373943f54c3ced
Submitter: "Zuul (22348)"
Branch: master

commit a13a27c767070770b17bf4f9eb373943f54c3ced
Author: Caio Bruchert <email address hidden>
Date: Thu Oct 13 13:30:10 2022 -0300

    Debian: fix VLAN creation in duplex-direct mode

    In duplex-direct mode, the ifupdown VLAN configuration tries to
    disable DAD using pre-up command 'sysctl -wq net.ipv6.conf.%s.accept_dad=0'
    where %s is the vlan device and fails because the VLAN is not created yet.

    This happens because Debian's ifupdown tool runs pre-up commands from
    /etc/network/interface.d/ before running /etc/network/if-pre-up/vlan
    script that creates the VLAN.

    Moving the sysctl command to up or post-up won't avoid DAD to run.

    The solution found is to add a VLAN creation command in pre-up before
    the sysctl command.

    Some bonding configuration was also being added to VLANs. Although it
    was not causing any issues, it was fixed to be added only to the
    bonding interfaces.

    Test Plan:
        PASS: Run new VLAN unit test for duplex-direct mode
        PASS: Install duplex-direct system

    Closes-Bug: 1992832

    Signed-off-by: Caio Bruchert <email address hidden>
    Change-Id: Icbdd113e6051d584959039e85f2dfa078ec0eb2c