Hello,
I am trying to understand the main differences between OVS bonding and
Linux bonding. I tried googling, but couldn't find much information. I have
few questions and appreciate your answers:
1. How is OVS bonding different from Linux bonding?
2. Is there any performance gain or loss of usin
discuss [mailto:discuss-boun...@openvswitch.org] *On Behalf Of
> *Nussbaum,
> Jacob
> *Sent:* Tuesday, December 2, 2014 9:59 PM
> *To:* 'discuss@openvswitch.org'
> *Subject:* [ovs-discuss] OVS Bonding only using 1 link
>
>
>
> Hello all,
>
> I have another
ber 2, 2014 9:59 PM
To: 'discuss@openvswitch.org'
Subject: [ovs-discuss] OVS Bonding only using 1 link
Hello all,
I have another bonding issue with OVS I was hoping someone could help with.
I have two virtual machines. Each with 5 network interfaces (eth0 -eth4).
Eth1-4 are bonded betwe
Hello all,
I have another bonding issue with OVS I was hoping someone could help with.
I have two virtual machines. Each with 5 network interfaces (eth0 -eth4).
Eth1-4 are bonded between the two VM's. The mininet VM has multiple "hosts"
that I have the traffic mirrored over the bonded port (b
On Tue, Oct 22, 2013 at 09:58:26PM +0530, ananthan wrote:
> Some vms are facing packet loss ~40% ,i am trying to understand why its
> happening,(*lost: *field is 0)
>
> As mentioned in the early mail,*used* counter for both drop action and vlan
> action is increasing:
>
> *in_port(2),*eth(src=00:2
thanks for the reply Ben,
Some vms are facing packet loss ~40% ,i am trying to understand why its
happening,(*lost: *field is 0)
As mentioned in the early mail,*used* counter for both drop action and vlan
action is increasing:
*in_port(2),*eth(src=00:26:0a:24:b0:00,dst=fe:34:37:a3:fd:9d),eth_typ
On Mon, Oct 21, 2013 at 10:25:50PM +0530, ananthan wrote:
> For some guest there is packet loss,can this be because of this drop action
> in eth0, its "*used"* value is also getting updated.
>
> in_port(2),eth(src=00:26:0a:24:b0:00,dst=ff:ff:ff:ff:ff:ff),eth_type(0x8100),vlan(vid=32,pcp=0),encap(e
Thanks Ben for the reply,
For some guest there is packet loss,can this be because of this drop action
in eth0, its "*used"* value is also getting updated.
in_port(2),eth(src=00:26:0a:24:b0:00,dst=ff:ff:ff:ff:ff:ff),eth_type(0x8100),vlan(vid=32,pcp=0),encap(eth_type(0x0806),arp(sip=15.x.x.130,tip
On Mon, Oct 21, 2013 at 07:27:59PM +0530, ananthan wrote:
> But Why there are flows for in_port(2) when bonding shows active-backup?
OVS can't control what packets come in on an interface.
> All the flows for in_port(2) shows action as drop,what is the need of these
> flows?
OVS has to do somet
Hi,
i have ovs bond bond0 with eth0 and eth1 connected to it
*ovs-appctl bond/list*
bond type slaves
bond0 active-backup eth0, eth1
*ovs-appctl bond/show bond0*
bond_mode: active-backup
bond-hash-basis: 0
updelay: 31000 ms
downdelay: 200 ms
lacp_negotiated: false
slave eth0: enabled
may_ena
Ben,
thanks, but my C skills is not so good to do this.
OVS way of bonding management is much more flexible than using OS-
dependent syntax of configuring in-kernel bonding driver.
--
*/Best regards,/*
/Eugene Istomin/
> On Fri, Sep 27, 2013 at 10:41:19AM +0300, Eugene Istomin wrote:
> > Do y
On Fri, Sep 27, 2013 at 10:41:19AM +0300, Eugene Istomin wrote:
> Do you have plans to get OVS bond working as single port with flows on
> it or we should use the linux bond?
We've considered various options over the years but it's never been a
big enough issue to actually come up with a solution.
Hello,
in the way of templating OVS rules for hypervisor we use approach like this:
/openflow_add_aspoof:/
ovs-ofctl add-flow $switch "in_port=$port priority=31000 cookie=$cookie
dl_type=0x0800 dl_src=$mac nw_src=$ip action=normal"
ovs-ofctl add-flow $switch "in_port=$port priority=31000
I doubt whether the MAC address of the bond interfaces is important.
Rather, I think that the source MAC address of the packets transmitted
over the bond is the important issue. This means that you would have to
rewrite the MAC addresses on the packets transmitted over the bond.
On Tue, Feb 05, 2
It means that can I make bonding with 2 wireless interfaces with same MAC
address?
On Tue, Feb 5, 2013 at 3:18 PM, Ben Pfaff wrote:
> On Tue, Feb 05, 2013 at 03:11:52PM +0900, Yuris Mulya Saputra wrote:
> > I have a problem to make OVS bonding using 2 Wireless interfaces as AP or
> > master mod
On Tue, Feb 05, 2013 at 03:11:52PM +0900, Yuris Mulya Saputra wrote:
> I have a problem to make OVS bonding using 2 Wireless interfaces as AP or
> master mode (from 2 APs to 1 AP).
> First I make a bridge as usual and then make bonding with 2 wireless
> interface. But when I make bonding, the inter
Hi All,
I have a problem to make OVS bonding using 2 Wireless interfaces as AP or
master mode (from 2 APs to 1 AP).
First I make a bridge as usual and then make bonding with 2 wireless
interface. But when I make bonding, the internet is diconnected for the
client. My 2 APs have different MAC addre
I think the difference here is that vPC, Multi-chassis Etherchannel, ... allows
load-balancing over all paths (over both switches), right?
While this surely provides HA, it comes with a cost for the Nexus-like switches
that it requires.
I was thinking about load-balancing over the same switch,
What you are asking for seems to be a vPC (virtual port channel, aka MLAG)
between switch 1 and switch2.
Thanks.
On Mon, Aug 6, 2012 at 8:02 AM, Frido Roose wrote:
> Hello,
>
> Is it possible to create an active-backup bond on top of a balance-slb
> bond in openvswitch?
>
> This is the situatio
You may consider using LACP to solve this problem. You can create a
single 4 port LACP bond, and plug it into two different switches.
LACP will automatically choose one of those two switches to be active
and forward to it until it dies.
Ethan
On Mon, Aug 6, 2012 at 8:02 AM, Frido Roose wrote:
Hello,
Is it possible to create an active-backup bond on top of a balance-slb bond
in openvswitch?
This is the situation:
1 bridge: br0
eth0 => physical switch1
eth1 => physical switch1 (same switch or stack)
eth2 => physical switch2
eth3 => physical switch2 (same or stack)
__
21 matches
Mail list logo
