On Mon, 2007-06-11 at 08:32 +0800, Leon Xiang Li wrote:
> When I run polymix-4, I meet the following problem:
>
> 1. In polygraph-2.8. when the requests/second reach around 50, the
> client will keep replying :
>
> 006.91| Xaction.cc:74: error: 1024/1024 (c19) unsupported HTTP status code
What was the unsupported HTTP status code? You may want to use the
"--dump errs" command line option to see the details of an error
message. Usually, Squid replies with an error if it is overloaded or
cannot reach the origin server. The response headers and body will have
more detail about the error.
> And in squid cache.log, there is error message.
What error message?
> 2. So I adjust the rr/s to 50, and rerun the test. after runing for a
> while, I can see the client issue request around 50 request per second
> (request rate), but in sever request rate is 0, and in squid, there is
> no requesting log in access.log, it seems that request do not send to
> squid at all. And polyclt do not report error too.
>
> Can you give me amy advice ?
I would need more information, especially if you are seeing that polyclt
is sending requests to Squid, receiving responses, but Squid does not
show any of that activity.
I can only imagine a similar situation if polyclt robot connections are
refused by the Squid box or if there is something between Polygraph and
Squid that refuses/aborts connections. In all those cases, all Polygraph
transactions will end with errors, which should be visible on the
console log.
HTH,
Alex.
> Alex Rousskov wrote:
> > On Fri, 2007-06-01 at 10:51 +0800, Leon Xiang Li wrote:
> >
> >> I had tried to setup the polymix-4 test as your suggestion.
> >>
> >> I run polysrv on 174.19.100.1
> >> I run polyclt on 174.19.100.2
> >> I run squid on 174.19.100.3
> >>
> >> all the three servers are RHEL4 Linux.
> >>
> >> squid works well.
> >>
> >> My polymix-4.pg is enclosed
> >>
> >>
> >> But the test stop with server timeout.
> >>
> >> The client and server log is enclosed.
> >
> > It looks like your robots cannot connect to the proxy or cannot get any
> > packets back from the proxy. Did you follow the HOWTO below? Did you
> > test client<->proxy and proxy<->server IP connectivity with ping
> > (specifying the right alias address as the source and destination IPs;
> > see Section 4.2)?
> >
> > http://www.measurement-factory.com/docs/PolyMix-4/
> >
> > You will need to adjust the HOWTO a little to make work on Linux, but
> > all the major steps of setting up the network and DNS must be done for
> > the test to work.
> >
> > While it is possible to simplify the workload (e.g., remove domain
> > names) but it is usually easier to start with defaults. PolyMix is a
> > realistic and, hence, complex test; running it by hand, especially for
> > the first time, is not trivial.
> >
> > HTH,
> >
> > Alex.
> >
> >> 003.43| Connection.cc:488: error: 1/1 (s110) Connection timed out
> >> 003.43| error: raw write after connect failed
> >> 003.43| connection between 10.101.0.246:37457 and 174.19.100.3:3128
> >> failed at 0 reads, 0 writes, and 1 xacts
> >
> >
> >
Received on Mon Jun 11 2007 - 12:25:38 MDT
This archive was generated by hypermail 2.2.0 : Tue Jun 12 2007 - 12:00:09 MDT