Note: This bug is displayed in read-only format because the product is no longer active in Red Hat Bugzilla.

Bug 1438517

Summary:	MTU setting on DPDK bonds
Product:	Red Hat OpenStack	Reporter:	Karthik Sundaravel <ksundara>
Component:	openvswitch-dpdk	Assignee:	Aaron Conole <aconole>
Status:	CLOSED NOTABUG	QA Contact:
Severity:	medium	Docs Contact:
Priority:	medium
Version:	12.0 (Pike)	CC:	aloughla, astupnik, fleitner, ksundara, nyechiel, skramaja, vchundur, ysubrama
Target Milestone:	---
Target Release:	---
Hardware:	Unspecified
OS:	Unspecified
Whiteboard:
Fixed In Version:		Doc Type:	If docs needed, set a value
Doc Text:		Story Points:	---
Clone Of:		Environment:
Last Closed:	2017-05-17 19:52:55 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:

Description Karthik Sundaravel 2017-04-03 15:38:36 UTC

Regarding jumbo frame support in DPDK ports and bonds, I need some clarity.

IIUC, for DPDK ports the MTU configuration is applied on the interface only. And it is not required to specify the mtu on the DPDK bridge. 

  ovs-vsctl add-port br-link dpdk0 -- set Interface dpdk0 type=dpdk -- set Interface dpdk0 mtu_request=9000

Please correct if my understanding is wrong.

When we create the DPDK Bond with below command, it is not clear on where and how to set the MTU value. Should we set the MTU value on both the interfaces dpdk0 and dpdk1 or we need it on bond?

  ovs-vsctl add-bond br-link dpdkbond0 dpdk0 dpdk1 -- set Interface dpdk0 type=dpdk -- set Interface dpdk1 type=dpdk

Comment 1 Vijay Chundury 2017-04-24 11:29:23 UTC

Karthik,
I think you should have a chat with the OVS team/Andrew and talk to them about handling mergeable buffers, keeping into context of when to turn-on/off of this feature when introducing the Jumbo setting.
It is learnt that QEMU mergeable buffers when turned off would bump up the perf numbers positively (recommended by OVS-DPDK community).
Ensure you get a clarity on end to end setting required for this feature from the installer perspective (HA as well).


Regards
Vijay

Comment 3 Aaron Conole 2017-05-17 18:32:59 UTC

(In reply to Karthik Sundaravel from comment #0)
> Regarding jumbo frame support in DPDK ports and bonds, I need some clarity.
> 
> IIUC, for DPDK ports the MTU configuration is applied on the interface only.
> And it is not required to specify the mtu on the DPDK bridge. 
> 
>   ovs-vsctl add-port br-link dpdk0 -- set Interface dpdk0 type=dpdk -- set
> Interface dpdk0 mtu_request=9000

This is correct.  OvS 'bridges' aren't the same as a linux bridge.  There's no concept of a datapath-wide mtu.

> When we create the DPDK Bond with below command, it is not clear on where
> and how to set the MTU value. Should we set the MTU value on both the
> interfaces dpdk0 and dpdk1 or we need it on bond?
> 
>   ovs-vsctl add-bond br-link dpdkbond0 dpdk0 dpdk1 -- set Interface dpdk0
> type=dpdk -- set Interface dpdk1 type=dpdk

You will need to set the MTU on all the interfaces to match, yes.

Comment 5 Karthik Sundaravel 2017-05-17 19:52:55 UTC

Thank Aaron.