On Dec 2, 9:11 am, Amit Kasher <[EMAIL PROTECTED]> wrote:
> Hi,
> Thanks for your reply. Answers are inline.
>
> On Dec 2, 5:50 pm, jchimene <[EMAIL PROTECTED]> wrote:> Hi,
>
> > A few questions:
>
> > o Are all packets sent to the server the same size?
>
> No, they are not.
>
> > o What is that size?
>
> This depends on the service call - somewhere between 150 and 2000
> bytes.
> I will mention again that by using a sniffer (tcpdump), it seems that
> EVERY time this issue occurs, the actual packets the server receives
> are ALWAYS EXACTLY 80% of what it should have received. This, again,
> was very encouraging to find as a clue, but unfortunately led me
> nowhere.

At this point, I'd write a ping script and start generating packets of
a certain size. Hammer the server to see if you can reproduce on
demand. If you can reproduce w/ ping, it's not a browser/server issue.

> > o Have you checked for other types of congestion?
>
> Congestion? Unfortunately, I don't have any control over the client's
> environment since this is an internet application and I can't
> reproduce it.

I don't mean the client congestion, I mean congestion en route, i.e.
the cloud between the client and the server. Your later answers seem
to eliminate The Cloud.

> > o Is this entirely TCP/IP? Have you checked maxrss?
>
> maxrss? I'm not sure I understood the relevance... TCP/IP is obviously
> used, it is the underlying protocol of HTTP...

My bad. I meant MTU. But that doesn't sound like it's relevant.

> > o Have you enabled logging on intermediate nodes to see if there are
> > congestion issues?
>
> I wish I could... I don't have any control over any node before the
> server. It is a CentOS VPS hosted internet application. I will state
> that this occurred in several hosting providers, in several countries
> and geographical locations.

So it's sounding more and more like the app. If it's several hosting
providers in several locations, that's pretty much A Clue.

> > o Is this related to a specific time of day (although it probably
> > happens between 10:00 and 14:00...)
>
> I didn't find any correlation between the time of day and the
> occurrence of this. Obviously, this is normalized to the usage load,
> as you implied.
>
> > o Do you have a world-wide net? If so, does the problem travel across
> > time zones?
>
> My users are not from around the world, but as I stated - this issue
> occurred when using hosting providers around the world.
>

OK, so we're down to the app. I'd try constructing a reproducer using
ping with specific packet sizes. Record the output stats and run them
through a formatting routine that will make it easier to check for
problems. The goal here is to check end-to-end transmission w/o using
application layer code. If this transmission failure happens every
day, you should see something happen within 24 hours.

If nothing untoward happens, it's (probably) the app. In that case,
I'd grab a copy of Perl (or whatever you want) to write a client-side
app that faithfully simulates your app's transmission profile. Point
it at one of your servers and stress test that server/client circuit.

Send PCs to several locations if you have to, but get something in the
field that can reproduce this independently of your app and over which
you have complete control. If that's not realistic, you'll need a
"lab" environment, i.e. a machine that isn't a developer box, can be
wiped clean and conveniently set to a known state.

Bueno Suerte,
jec

>
> > On Dec 2, 2:13 am, Amit Kasher <[EMAIL PROTECTED]> wrote:
>
> > > Hi,
> > > Does anyone has any new insights about this issue? We've been
> > > investigating for over a year(!), and we seem to not be the only
> > > ones...
>
> > >http://tinyurl.com/5rqfp5
>
> > > Thanks.
--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups 
"Google Web Toolkit" group.
To post to this group, send email to Google-Web-Toolkit@googlegroups.com
To unsubscribe from this group, send email to [EMAIL PROTECTED]
For more options, visit this group at 
http://groups.google.com/group/Google-Web-Toolkit?hl=en
-~----------~----~----~----~------~----~------~--~---

Reply via email to