> -----Original Message-----
> From: Testbed Support for GridPP member institutes [mailto:TB-
> [log in to unmask]] On Behalf Of Stuart Purdie
> Sent: 30 June 2010 17:02
>
> Therefore, the advice from me is to set "net.ipv4.tcp_sack = 1" for
the
> worker nodes behind NAT.
>
Given that this phenomenon is somewhat timing dependent, would it be
reasonable to describe it as a sack race condition?
Ewan
|