Re: [PATCH] tcp_cubic: limit delayed_ack ratio to prevent divide error

[TB <lkml@techboom.com>]

On 11-05-06 12:53 PM, Stephen Hemminger wrote:
> On Fri, 06 May 2011 12:15:46 -0400
> TB <lkml@techboom.com> wrote:
> 
>> -----BEGIN PGP SIGNED MESSAGE-----
>> Hash: SHA1
>>
>> On 11-05-04 04:53 PM, Brandeburg, Jesse wrote:
>>>
>>>
>>> On Wed, 4 May 2011, Stephen Hemminger wrote:
>>>
>>>> TCP Cubic keeps a metric that estimates the amount of delayed
>>>> acknowledgements to use in adjusting the window. If an abnormally
>>>> large number of packets are acknowledged at once, then the update
>>>> could wrap and reach zero. This kind of ACK could only
>>>> happen when there was a large window and huge number of
>>>> ACK's were lost.
>>>>
>>>> This patch limits the value of delayed ack ratio. The choice of 32
>>>> is just a conservative value since normally it should be range of 
>>>> 1 to 4 packets.
>>>>
>>>> Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
>>>
>>> patch seems fine, but please credit the reporter (lkml@techboom.com) with 
>>> reporting the issue with logs, maybe even with Reported-by: and some kind 
>>> of reference to the panic message or the email thread in the text or 
>>> header?
>>
>> We're currently testing the patch on 6 production servers
> 
> Thank you, is there some regularity to the failures previously?

Not really, there was more chance of it happening after a reboot and
during the night (when there is less traffic) for some weird reason.

As a workaround we switched most of the servers to reno