Created attachment 1582629 [details]
the same journal for rhel 8.0

Reassigning to NetworkManager for debugging/info/hints. Please see the log in the Description (failing 8.1 nightly) and comment #1 (working 8.0).

Hi,

I tried on RHEL 8.0 and it breaks when upgrading libteam from 1.27-10 to 1.28-1. The cause of the error is that teamd_config_port_set() fails to find the port by ifname.

Xin, do you have any idea why?

In NetworkManager we first send the port configuration and then actually enslave the port interface to the team master. Perhaps teamd can't find the port because the interface is not yet enslaved. But this used to work in the past.

(In reply to Beniamino Galvani from comment #3)
> Hi,
> 
> I tried on RHEL 8.0 and it breaks when upgrading libteam from 1.27-10 to
> 1.28-1. The cause of the error is that teamd_config_port_set() fails to find
> the port by ifname.
> 
> Xin, do you have any idea why?
> 
> In NetworkManager we first send the port configuration and then actually
> enslave the port interface to the team master. Perhaps teamd can't find the
> port because the interface is not yet enslaved. But this used to work in the
> past.

teamd_config_port_set() was introduced on rhel-8.1 by:

  https://github.com/jpirko/libteam/commit/c8b356a3cd

Hangbin, can you help take a look at it? thanks.

Reproducer:

# nmcli connection add type team ifname team0 con-name team0+ team.config '{"runner": {"name": "activebackup"}}' ip4 172.20.1.3/24
Connection 'team0+' (b95f23f0-2da4-41bd-969a-df228c8f056e) successfully added.

# nmcli connection add type ethernet ifname enp8s0 master team0 con-name slave+ team-port.config '{"prio": -10, "sticky": true}'
Connection 'slave+' (09221211-9e06-4f75-9b6c-d003f63f0c82) successfully added.

# nmcli connection up slave+
Error: Connection activation failed: Unknown error

# journalctl --since '1 minute ago' | grep team
Jun 26 07:55:10 localhost.localdomain NetworkManager[1021]: <info>  [1561550110.9568] device (enp8s0): released from master device team0
Jun 26 07:56:08 localhost.localdomain NetworkManager[1021]: libteamdctl: cli_usock_process_msg: usock: Error message received: "ConfigUpdateFail"
Jun 26 07:56:08 localhost.localdomain NetworkManager[1021]: libteamdctl: cli_usock_process_msg: usock: Error message content: "Failed to update config."
Jun 26 07:56:08 localhost.localdomain NetworkManager[1021]: <error> [1561550168.8452] device (team0): failed to update config for port enp8s0 (err=-22)
Jun 26 07:56:08 localhost.localdomain NetworkManager[1021]: <info>  [1561550168.8457] device (enp8s0): released from master device team0

# rpm -q libteam
libteam-1.28-2.el8.x86_64

# grep PRETTY_NAME /etc/os-release
PRETTY_NAME="Red Hat Enterprise Linux 8.1 Beta (Ootpa)"

Thanks, Beniamino

teamd supports for ports config going first, then ports enslaving. The commit in c4 breaks it, I will make a change in teamd_config_port_update():

        tdport = teamd_get_port_by_ifname(ctx, port_name);
        if (tdport)
                err = teamd_config_port_set(ctx, tdport, port_new_obj);

Hi Vladimir,

I noticed that in the NM CI suite there isn't any test setting the 'team-port.config' property of team slaves... could we add at least one to have this feature covered? Thanks!

*** Bug 1719110 has been marked as a duplicate of this bug. ***

I verified this using reproducer in https://bugzilla.redhat.com/show_bug.cgi?id=1719110#c0 on libteam-1.28-4.el8.x86_64.

Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.

For information on the advisory, and where to find the updated
files, follow the link below.

If the solution does not work for you, open a new bug report.

https://access.redhat.com/errata/RHBA-2019:3550