|
Sorry for the late reply, I had to move our server park to a new
location. I will do some tests this week as I was able to remove traffic
from one of the load balancers.
Graeme Fowler wrote:
> On Mon, 2008-03-24 at 13:14 +0100, bgs wrote:
>> What can be the difference? In my setup moving vrrp to a physical eth
>> device solved the bind error.
>
> As I would expect.
>
>> Might be a bonding driver difference? What do you use?
>
> I've used the bond device before with no errors, but I wasn't using the
> load-balancing option on the bond interface. My issue was failover
> (spanning tree), so I had dual-homed the servers at layer 2 and made
> their "left hand" port the active part of the bond interface. Rebooting
> the switch they all connected to (or pulling its' power) made everything
> go out the "right hand" port instead.
>
> Interestingly, the hash mode you use is documented as follows:
>
> This algorithm is not fully 802.3ad compliant. A
> single TCP or UDP conversation containing both
> fragmented and unfragmented packets will see packets
> striped across two interfaces. This may result in out
> of order delivery. Most traffic types will not meet
> this criteria, as TCP rarely fragments traffic, and
> most UDP traffic is not involved in extended
> conversations. Other implementations of 802.3ad may
> or may not tolerate this noncompliance.
>
> I wonder if this is the cause of your problem? The key exchange could
> well result in TCP fragments, especially if you're using a large key.
>
> Try changing that to "layer2", and try changing the bond mode to
> "active-backup", just to see if it makes a difference.
>
> Graeme
>
>
> _______________________________________________
> LinuxVirtualServer.org mailing list - lvs-users@xxxxxxxxxxxxxxxxxxxxxx
> Send requests to lvs-users-request@xxxxxxxxxxxxxxxxxxxxxx
> or go to http://lists.graemef.net/mailman/listinfo/lvs-users
>
|
