Bug 349198

Summary:	=sys-apps/baselayout-1* missing slave in bond shouldn't be fatal
Product:	Gentoo Linux	Reporter:	Marcin Mirosław <bug>
Component:	[OLD] baselayout	Assignee:	Gentoo's Team for Core System packages <base-system>
Status:	RESOLVED TEST-REQUEST
Severity:	normal	CC:	floppym
Priority:	High
Version:	unspecified
Hardware:	All
OS:	Linux
Whiteboard:
Package list:		Runtime testing required:	---

Description Marcin Mirosław 2010-12-20 15:24:41 UTC

I have configuration in /etc/conf.d/net as follow:
slaves_bond0="eth0 eth1"
config_bond0=( "1.2.3.4/24" )

Everything is ok when both net devices exists in system. Now, let's imagine that eth1 became broken[1] and we didn't know about it. Network is still working, now i have to restart server (or network).
I got:
...
* Adding slaves to bond0
*   eth0 eth1
*   network interface eth1 doesn not exist
*   Please verify hardware or kernel module (diver)    [ !! ]

In this moment, i have at lest one eth working and available but my network is down. My server is not accessible. Now i have problem.
Behavior while bond is creating should be similar to behavior e.g RAID1, raid is assembled when at least one disk is present.

[1] broken in fatal way, hardware is not detected by system

Reproducible: Always

Comment 1 Mike Gilbert gentoo-dev

2010-12-21 05:23:00 UTC

I'm not sure that there is any way to detect the difference between a hardware failure and a configuration error here.

Network bonding also seems quite different from RAID in that network adapters fail far less frequently than hard disks.

Comment 2 Robin Johnson archtester

2010-12-21 06:45:55 UTC

Please retest with the latest git of openrc, where this has changed.

Comment 3 Marcin Mirosław 2010-12-21 09:38:10 UTC

(In reply to comment #1)
> I'm not sure that there is any way to detect the difference between a hardware
> failure and a configuration error here.

I have no idea how to detect difrence too. Imho it's nothing to worry about. For me, ewarn is enough to warn user about missing device.
 
> Network bonding also seems quite different from RAID in that network adapters
> fail far less frequently than hard disks.

Agree. It was analogy.

(In reply to comment #2)
> Please retest with the latest git of openrc, where this has changed.

For me nothing changed between openrc-0.6.8 and openrc-9999 - bond device is crated, even without any slave present. Both version works correctly for me.
And what about stable baselayout-1.x.x? Is any chance to fix it?

Comment 4 Robin Johnson archtester

2010-12-21 09:46:22 UTC

No, we're not backporting to baselayout1. I'm trying to get all oldnet issues solved with it so we can stabilize.