Re: cannot connect to http server

Markus Heinz wrote:

on Windows XP. It has worked like expected.
>
> Test 2: I have accessed http://home.arcor.de/markusheinz/index.html with
> Konquerer (KDE 3.1.4) and linux 2.6.1 from a Knoppix CD (Knoppix c't
> edition 3.4). The browser said "Waiting for home.arcor.de..." and
> nothing happened. After a long time it showed a "Connection timed out".
> The ethereal dump shows that it did not even send out a http request.
> Instead it sent four TCP packets with ACK and an old sequence number.
>

I'm using SuSe 8.2 Linux 2.4 Konqueror KDE 3.1.1 and it works accessing
your html...

Hi Jerry,

Jerry Smiley wrote:
>
>
> Markus Heinz wrote:
>
> on Windows XP. It has worked like expected.
>>
>> Test 2: I have accessed http://home.arcor.de/markusheinz/index.html with
>> Konquerer (KDE 3.1.4) and linux 2.6.1 from a Knoppix CD (Knoppix c't
>> edition 3.4). The browser said "Waiting for home.arcor.de..." and
>> nothing happened. After a long time it showed a "Connection timed out".
>> The ethereal dump shows that it did not even send out a http request.
>> Instead it sent four TCP packets with ACK and an old sequence number.
>>
>
> I'm using SuSe 8.2 Linux 2.4 Konqueror KDE 3.1.1 and it works accessing
> your html...

It didn't work for me about an hour ago, but now it does. Seems to be
something temporary...

LLAP, Martin

Martin
While it is working try tracing the packets again (tcpdump / ethereal) for a
baseline to compare against if it fails again...Also, maybe check the
ethernet card ipconfig for errors when the failure occurs..

> Hi Jerry,
>
> Jerry Smiley wrote:
>>
>>
>> Markus Heinz wrote:
>>
>> on Windows XP. It has worked like expected.
>>>
>>> Test 2: I have accessed http://home.arcor.de/markusheinz/index.html with
>>> Konquerer (KDE 3.1.4) and linux 2.6.1 from a Knoppix CD (Knoppix c't
>>> edition 3.4). The browser said "Waiting for home.arcor.de..." and
>>> nothing happened. After a long time it showed a "Connection timed out".
>>> The ethereal dump shows that it did not even send out a http request.
>>> Instead it sent four TCP packets with ACK and an old sequence number.
>>>
>>
>> I'm using SuSe 8.2 Linux 2.4 Konqueror KDE 3.1.1 and it works accessing
>> your html...
>
> It didn't work for me about an hour ago, but now it does. Seems to be
> something temporary...
>
> LLAP, Martin

Hi Jerry,

Jerry Smiley wrote:
> Martin
> While it is working try tracing the packets again (tcpdump / ethereal) for a
> baseline to compare against if it fails again...Also, maybe check the
> ethernet card ipconfig for errors when the failure occurs..

It seems to fail sometimes and sometimes not.

I uploaded traces of both cases:

http://mh57.de/~martin/trace.bad.dmp
http://mh57.de/~martin/trace.ok.dmp

Both first answering packet have a different TTL (56 vs 247) and the
already mentioned Window Size difference.

Maybe some broken load balancer?

LLAP, Martin

Martin Hermanowski <usenet2004@tramspap.mh57.de> wrote:


> It seems to fail sometimes and sometimes not.
>
[traces]
> Both first answering packet have a different TTL (56 vs 247) and the
> already mentioned Window Size difference.
>
> Maybe some broken load balancer?

I guess it's one broken system among the systems from which the
load balancer selects one. In the working case (the one without
the zero window) the webserver identifies itself as
Apache/1.3.26 (Unix) FrontPage/5.0.2.2623 AuthMySQL/2.20

Could anyone with Windows running, check what the server
indentificaion is in the zero window case? May be the $OP?


Ciao, Horst
--
»When pings go wrong (It hurts me too)« E.Clapton/E.James/P.Tscharn

Martin...
Looking at the traces it could be a network related problem where a Lan /
router may be causing the TTL to be set incorrectly...I see the 56 vs 247
TTL as well. This maybe causing the problem...

When it fails again...try ping the host and see if the TTL changes...
>Martin Hermanowski wrote:
>
> Both first answering packet have a different TTL (56 vs 247) and the
> already mentioned Window Size difference.
>
> Maybe some broken load balancer?
>
> LLAP, Martin

Martin

When I ping from my network, I'm getting a TTL 238 vs 247 when it fails in
your network...

64 bytes from home.arcor-online.net (151.189.20.30): icmp_seq=81 ttl=238
time=123 ms
64 bytes from home.arcor-online.net (151.189.20.30): icmp_seq=82 ttl=238
time=124 ms


Martin Hermanowski wrote:

> Hi Jerry,
>
> Jerry Smiley wrote:
>> Martin
>> While it is working try tracing the packets again (tcpdump / ethereal)
>> for a baseline to compare against if it fails again...Also, maybe check
>> the ethernet card ipconfig for errors when the failure occurs..
>
> It seems to fail sometimes and sometimes not.
>
> I uploaded traces of both cases:
>
> http://mh57.de/~martin/trace.bad.dmp
> http://mh57.de/~martin/trace.ok.dmp
>
> Both first answering packet have a different TTL (56 vs 247) and the
> already mentioned Window Size difference.
>
> Maybe some broken load balancer?
>
> LLAP, Martin

Hello,

On Sat, 14 Feb 2004 17:55:45 +0100
Horst Knobloch <horschti2@gmx.de> wrote:

> I guess it's one broken system among the systems from which the
> load balancer selects one. In the working case (the one without
> the zero window) the webserver identifies itself as
> Apache/1.3.26 (Unix) FrontPage/5.0.2.2623 AuthMySQL/2.20
>
> Could anyone with Windows running, check what the server
> indentificaion is in the zero window case? May be the $OP?

The server identification is the same in the zero window case.

Regards,

Markus

--
Visit my homepage: http://home.arcor.de/markusheinz/

Markus Heinz <su1690@studserver.uni-dortmund.de> wrote:

> On Sat, 14 Feb 2004 17:55:45 +0100
> Horst Knobloch <horschti2@gmx.de> wrote:
>
>> I guess it's one broken system among the systems from which the
>> load balancer selects one. In the working case (the one without
>> the zero window) the webserver identifies itself as
>> Apache/1.3.26 (Unix) FrontPage/5.0.2.2623 AuthMySQL/2.20
>>
>> Could anyone with Windows running, check what the server
>> indentificaion is in the zero window case? May be the $OP?
>
> The server identification is the same in the zero window case.

However even if it identifies itself the same, it's still
possible that one of the webservers is broken. I can't see
any way to find out more information from the outside. I
think you should talk to your ISP. Something (webserver,
load balancer, etc.) on their side is broken.

BTW: At the moment I can't reproduce the error.


Ciao, Horst
--
»When pings go wrong (It hurts me too)« E.Clapton/E.James/P.Tscharn