Dropped connections with tcp_tw_recycle=1

satish M G tcpipqa at gmail.com
Tue Jan 17 11:46:06 CET 2012


I am new to this list.
In continuation to the thread "Dropped connections with tcp_tw_recycle=1" I
have a query -

we faced the same issue of dropping the connections with tcp_tw_recycle set
and NAT.

we then tried with tcp_tw_reuse and it worked fine with the NAT and large
number of incoming connections.

however the documentation says that tcp_tw_reuse -

tcp_tw_reuse - BOOLEAN465
<http://www.mjmwired.net/kernel/Documentation/networking/ip-sysctl.txt#465>		Allow
to reuse TIME-WAIT sockets for new connections when it is466
<http://www.mjmwired.net/kernel/Documentation/networking/ip-sysctl.txt#466>		safe
from protocol viewpoint. Default value is 0.467
<http://www.mjmwired.net/kernel/Documentation/networking/ip-sysctl.txt#467>		It
should not be changed without advice/request of technical468
<http://www.mjmwired.net/kernel/Documentation/networking/ip-sysctl.txt#468>		experts.

The 'advice/request' part is to explain that:

 - Connections that go through TCP state aware nodes, such as firewalls, NAT

   devices or load balancers may see dropped frames.  The more connections

   there are, the more likely you will see this issue.

My query is -

1) does tcp_tw_reuse work well with large number of incoming connections
and NAT device between the client and server?

2) if tcp_tw_reuse does not work well, which part of the linux code is
causing the dropped frames?

3) generally what is the difference between tcp_tw_reuse and tcp_tw_recycle?


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://www.varnish-cache.org/lists/pipermail/varnish-misc/attachments/20120117/96e661f3/attachment.html>

More information about the varnish-misc mailing list