On 10/19/2012 7:37 AM, Edward Ned Harvey (lopser) wrote:
From: [email protected] [mailto:[email protected]]
On Behalf Of Andrew Hume
on the external VPN, we have 80 inbound feeds (10 per nodes) typically
around 23 Mbps each.
what we notice is that these socket connections occasionally go dry, that is,
data stops coming.
using tcpdump and sniffers, we determine that was because the server
starts sending
window size 0 messages back to the source systems. in fact, a sniffer
revealed
the window size starting around 26K and then quite quickly dropping all the
way down to 10, 8, 1 and then zero.
Apply your system updates. Make sure you're running the biggest best NICs available - there's a reason why they sell
NICs for "desktop" and for "server" even though you might think "what's the difference?"
And of course, they sell various levels of "server" nic. Get the biggest best ones.
Apply firmware updates, and look for updated drivers from your mfgr.
If all your boxes are a single brand and a single model, try injecting some
other brands and models.
The behavior you described isn't supposed to happen. You're experiencing a bug in
something low-level. You might be able to simply make the problem go away with
system& firmware updates, or swapping small(sic) pieces of hardware like the
NIC.
As an aside, the SolarFlare 10G cards are very, very nice. Built-in TCP
offloading, also interrupt coalscing to lower the burden on the host,
and interrupt balancing across cores. The amount of time it will wait
before passing up packets to the OS is also configurable as well as a
bunch of tunables. They work well on Solaris or Linux. They are
reasonably priced, and the first cards I've gotten > 9gbit/sec on for a
10gbit interface. Highly recommended vs. multiple 1gig built-ins.
_______________________________________________
Tech mailing list
[email protected]
https://lists.lopsa.org/cgi-bin/mailman/listinfo/tech
This list provided by the League of Professional System Administrators
http://lopsa.org/
