Hi Willy,

Thank you for your reply. And thank you guys for a great product.

The requests are not coming from a single or even a few IP addresses and
that what make us worry a bit though we have not got any complains from our
users about errors or anything similar.

I also tried to filter out those requests with ACL checking for only 1 byte
in the request but with no success.
Did you ever occur such a requests? (NULL byte)

Best Regards,
Shay


On Wed, Dec 5, 2012 at 11:47 PM, Willy Tarreau <[email protected]> wrote:

> Hi,
>
> On Thu, Nov 29, 2012 at 12:34:46AM +0200, SBD wrote:
> > Hi,
> >
> > We have problem on our production server which having a big amount of
> what
> > appear to be null requests (about 1K/min). We add the option dontlognull
> to
> > our frontends but they still being logged and display as request errors
> in
> > the stats page.
> >
> > We are using the last stable version (1.4.22). This is an example of what
> > we get when running "show errors":
> >
> > frontend ****** (#1): invalid request
> >   src ******, session #3468, backend <NONE> (#-1), server <NONE> (#-1)
> >   HTTP internal state 26, buffer flags 0x00909002, event #0
> >   request length 1 bytes, error at position 0:
> >
> >   00000  \x00
> >
> > As you can see the request is not empty but rather include one byte which
> > is NULL. I'm not sure if this is what called "NULL Reqeust" but we have
> no
> > idea what causes this (except for port scanning or something similar)
>
> No, as indicated in the doc :
>
>    it is possible to enable option "dontlognull" to indicate that a
>    connection on which no data has been transferred will not be logged
>
> Here you have a data transfer since the client sends one byte. so it is
> normal that there is a log. Can't you simply block the source IP address ?
> For example, using "tcp-request connection reject if { src XX.XX.XX.XX }" ?
>
> This should avoid the logs as the connection will be dropped just after
> the accept(). Just an idea. Or maybe it's an internal component which
> needs to be fixed ? After all, sending such a byte on an HTTP port is
> clearly an indication of someone looking for trouble.
>
> Regards,
> Willy
>
>

Reply via email to