Bug 1264029

Summary:	[Doc] Bonding modes 0, 5 and 6 should be avoided for VM networks - bring in from 6.7 docs
Product:	Red Hat Enterprise Linux 7	Reporter:	Laura Novich <lnovich>
Component:	doc-Virtualization_Deployment_and_Administration_Guide	Assignee:	Platform Virt Docs <virt-docs>
Status:	CLOSED CURRENTRELEASE	QA Contact:
Severity:	high	Docs Contact:
Priority:	high
Version:	7.2	CC:	ailan, bmcclain, byount, chuffman, cww, danken, ecs-bugs, iheim, jbainbri, jbuchta, jhradile, kernel-qe, knoel, lnovich, lpeer, michele, mkalinin, mleitner, mst, nyechiel, pep, peterm, rbalakri, Rhev-m-bugs, rkhan, sputhenp, swadeley, vyasevic, yeylon
Target Milestone:	rc	Keywords:	Documentation, Triaged
Target Release:	---
Hardware:	All
OS:	Linux
Whiteboard:	network
Fixed In Version:		Doc Type:	Bug Fix
Doc Text:		Story Points:	---
Clone Of:	1097143	Environment:
Last Closed:	2015-12-20 08:40:12 UTC	Type:	Bug
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:
Embargoed:
Bug Depends On:	1097143, 1097312
Bug Blocks:	1075802, 1095589, 1173188, 1286552

Comment 5 Dan Kenigsberg 2015-11-17 11:09:37 UTC

(In reply to Jamie Bainbridge from comment #4)

> > Under no circumstances should Modes 0, 3, 5, or 6 be used.
> 
> I assume this comes from the RHEV documentation?
> 
> There is some discrepancy between what RHEV decided to allow for guest
> networks and what actually works. The knowledgebase solution I wrote
> (https://access.redhat.com/solutions/67546) lists 0 and 3 as working, a view
> which former bonding maintainer vfalico supported (Bug 1097312).

This discrepancy should be fixed on RHEV side. IICR vfalico and jpirko asked to block modes 0 5 6 when bridges are connected when discussion multiple customer tickets due their usage.

If mode 0 can actually be enabled (with caveats) by the platform, I suppose RHEV should enable it as well.

Karen, is this the current wisdom? mode 0 is good for us?

Comment 6 Karen Noel 2015-11-17 11:15:38 UTC

Amnon's networking team will be able to answer. Thanks.

Comment 7 Vlad Yasevich 2015-11-19 16:19:52 UTC

WRT to bond mode 0:

Bonding mode-0 requires manual configuration on the switch that all bond members
are connected to.  The switch configuration requires all ports to be added to a single LAG (or EtherChannel) because the bond will use the same source mac on
each bond member, and without LAG on the switch, the flow will bounce between multiple ports causing unreachability.

This requirement for manual switch configuration is the reason that we do not recommend bond mode 0 being used.

-vlad

Comment 9 Vlad Yasevich 2015-11-24 18:32:17 UTC

The changes wrt to bonding modes look fine to me.

Comment 11 Laura Novich 2015-12-20 08:40:12 UTC

Published on December 17 version 2-24

Comment 12 Laura Novich 2015-12-20 08:43:06 UTC

Published on December 17 version 2-24