Some of the connection reset errors are unavoidable when the
workload includes HTTP persistent connections due to HTTP operation.
However, the number of those errors should be negligible.
If the number of errors is small, what you may want check first is
that the errors are indeed caused by racing conditions on persistent
connections. To do that, disable persistent connections by setting
agents' conn_use_lmt distribution to "const(1)". If most errors go
away, and there were just a few of them to start with, you have
nothing to worry about.
Otherwise, there is probably a problem with your network setup or
kernel configuration. Things you may want to double check:
- L2 connectivity is good
See "Section 4" (but delete dummynet rules for these tests!):
- You reconfigured the kernel after installing our iso image.
See "Kernel Config File" section at
- Polygraph was rebuilt from scratch after the kernel was
reconfigured (make distclean; ./configure; make)
If you are still unsatisfied with what you see, please send me
complete client and server console output (cut after you get lots of
errors logged and compressed if needed), and we will go from there.
On Wed, 25 Apr 2001, WESTOVER,ZANE (HP-USA,ex1) wrote:
> During the no proxy test using the polymix-3-guts.pg file, we keep getting
> the following error:
> ./Xaction.cc:79:error: (268)
> connection closed before sending headers
> which is followed by:
> ./connection.cc:175:error: (54)
> Connection reset by peer
> Everything seems to be fine until we reach the i-inc1 phase after 09.20 and
> it hits 70.99 connections/second.
> The lab consists of six(6) Pentium III 800 pc's with 128 MB of RAM each.
> They are connected by an HP ProCurve 2524 10/100 switch.
> We are using FreeBSD 3.4 (the iso was downloaded from your page) with the
> kernel modifications suggested on your page.
> I greatly appreciate any help you can offer.
> Thank you,
This archive was generated by hypermail 2b29 : Tue Jul 10 2001 - 12:00:19 MDT