[ovs-discuss] XenServer vSwitch 0.90.6

Jesse Gross jesse at nicira.com
Thu Nov 5 03:31:02 UTC 2009


Hi Danny,

Thanks for pointing this out.  There is a bug where once the updelay has 
begun it is not canceled, even if all the other slaves are disabled.  
There was an earlier, incomplete fix which would skip the updelay if a 
slave is brought up and there are no active slaves.

I just sent out a patch for review that will make the scenario you 
described work as expected.  I anticipate that it will make it into the 
development branch shortly and will be included in the 0.90.7 release.

Jesse

Danny Wannagat wrote:
> Hi.
>
> I'm currently testing bonding on XenServer 5.5 with vswitch 0.90.6
>
> When the MII Status of one of the bond interfaces is DOWN, it doesn't  
> switch to UP before the timout if the other interfaces goes down.
>
> Example:
> 1 Bond with ETH0 and ETH1.
> Both are active.
> MII Status UP
>
> Now I pull the network cable on ETH0. All vms switch to ETH1. (cat / 
> proc/net/bonding/bond0)
> MII Status ETH0: DOWN.
>
> Now I put the cable back into ETH0.
>
> 2 seconds later I pull ETH1. Now both interfaces are down.
>
> It takes until the updelay timeout runs out before eth0 gets online  
> again and I get access to the VMs.
>
> Without vswitch (in normal xenserver bonding). The DOWN interface  
> would immediately come online if I pull the cable from nic2.
> Before the updealy timeout runs out.
>
>
> Does that work as expected?
>
> Thanks
> Dan
>
> _______________________________________________
> discuss mailing list
> discuss at openvswitch.org
> http://openvswitch.org/mailman/listinfo/discuss_openvswitch.org
>   




More information about the discuss mailing list