RE: [FlashComm] Clients getting disconnected from FCS after idle
time 2004-02-26 - By Edward Chan
Back
I just want to be clear that we're talking about the same thing. It sounds
like Graeme is talking about clients becoming "unresponsive" whereas it
sounds like Brian was talking about clients getting "disconnected". Brian,
is that correct, or did I misunderstand?
As for the "unresponsive" clients...we have addressed bugs related to this
in the patch. However, the problem we've addressed seems to be slightly
different than what Graeme describes. The problem we've fixed relates to
cleaning up "zombied" connections; connections that FCS thinks are still
there but really are not. This can happen for various reasons; abnormal app
shutdown, network failure, etc. What happens is that the connection on the
client end has been closed. However, FCS has not yet been notified. This
can happen because sometimes the OS doesn't close the socket right away, or
whatever. In any case, the connection looks valid on the server end. So
what FCS does, is it tries to ping the client. If it gets a reply, it knows
the client is still alive. If it doesn't get a reply, the client must be
dead, and we will explicitly close the connection on our end. This client
gc and ping mechanism is fixed in the up coming patch.
However, what Graeme describes, and I haven't gone back thru the list
archives to check, but it seems he describes a situation where a client is
connected, but not doing anything. Then after some time, it appears the
client is hosed; not disconnected, but unresponsive. Is that correct?
Actually, one thing that just popped into mind regarding mysteriously
disconnected clients is, I think firewalls or load balancers may be
configured to close connections if there has not been network activity for
some time. That may be happening in Brian's case.
Anyway, getting back to unresponsive clients...I'll try the simple app that
Graeme suggests. We have had apps with long lived client connections that
seem fine. But they are either doing something, or doing nothing, but in
both cases behaved as expected and did not get disconnected unless there was
a network error or something. I'll try letting a client be idle for some
time and then see if it responds to commands after that.
Ed
> -- --Original Message-- --
> From: Jorge Maiquez [mailto:j_maiquez@(protected)]
> Sent: Thursday, February 26, 2004 3:15 AM
> To: flashcomm@(protected)
> Subject: RE: [FlashComm] Clients getting disconnected from
> FCS after idle time
>
>
> I first noticed it ages ago (FCS1.0 I think) in apps where no
> connectionlight or any other pinging mechanism was being
> used. Haven't really noticed it since, but then again most of
> my apps now make use of some sort of pinging..
> No big help, I know.. just a confirmation that I've also seen
> this behaviour before, and I think it happened more with
> mac's than with pc's, but don't hold me to that :)
>
> Cheers,
> -Jorge
>
>
> Jorge Maiquez
> Technical Architect
>
> Digital Samba
>
> C. Viladomat 108, Pral. 1a
> 08015 Barcelona
> Tel.: +34 93 451 56 62
> Mobile: +34 647 236 027
> http://www.digitalsamba.com
>
> This e-mail is intended for the named addressee only. It may
> contain confidential and/or privileged information. If you
> have received this message in error, please let us know and
> then delete this message from your system. You should not
> copy the message, use it for any purpose or disclose its
> contents to anyone.
>
> -- --Original Message-- --
> From: graeme.bull@(protected)
> [mailto:graeme.bull@(protected)]
> Sent: Thursday,
> February 26, 2004 7:51 AM
> To: flashcomm@(protected)
> Subject: RE: [FlashComm] Clients getting disconnected from
> FCS after idle time
>
>
> Hi Ed,
>
> You're on fire or something! right on. This is an old bug
> and most people have got around it by using their own
> "pinging" function or something.
>
> I suggest just making a very simple application where you can
> press a button and the server side will return a value and
> show it in a text field. Build, run it, test it and then
> leave it for about... 2 mins or so and press that button
> again. It's not so much that the client gets disconnected,
> it's that the client on longer can send or receive with the
> server. Very strange.
>
> (we have tried the maxappidle time too, see the archives if
> you need confirmation, it's all there)
>
> Graeme
>
> -- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ------
> Supported by Fig Leaf Software
> -- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ------
> Lower Prices for Certified Training! Check out new lower
> prices for Certified Macromedia Training from Fig Leaf
> Software. Expand your skill set with courses in ColdFusion,
> Flash, Rich Internet Applications and .NET in the new year.
> Fig Leaf Software provides
> the highest caliber instruction at our training centers in
> Washington
> D.C., Atlanta, Chicago, Baltimore, Northern Virginia, or
> on-site at your location.
>
> Get the details at http://training.figleaf.com/
> -- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ------
> To unsubscribe, e-mail: flashcomm-unsubscribe@(protected)
> For additional commands, e-mail: flashcomm-help@(protected)
>
-- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ------
Supported by Fig Leaf Software
-- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ------
Lower Prices for Certified Training! Check out new lower prices for
Certified Macromedia Training from Fig Leaf Software. Expand your
skill set with courses in ColdFusion, Flash, Rich Internet
Applications and .NET in the new year. Fig Leaf Software provides
the highest caliber instruction at our training centers in Washington
D.C., Atlanta, Chicago, Baltimore, Northern Virginia, or on-site at
your location.
Get the details at http://training.figleaf.com/
-- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ------
To unsubscribe, e-mail: flashcomm-unsubscribe@(protected)
For additional commands, e-mail: flashcomm-help@(protected)
|
|
