issues on gws 1.9.18d

Hi all,


after an update to gws 1.9.18d, from 1.9.16 i have noticed
some issue that i want to share with you


1)
when i do a request for
a certain page (that should return a big xml) i got
<fancy xml>

<blabla/>
<blabla/>
<blabla/>
Http header (!?!?!)
<blabla/>
<blabla/>

<blabla/>
</fancy xml>

this bug was already present in the past, and
if i remember well was a problem with http1.0 or something like that (i
don't know if this is till the case)
In any case, what's happen is that
the header are sended after a piece of request is issued :(


2)
wget
issue
if i do a
prometeo:~# wget -S -v http://192.168.2.114:
2480/base_c2.swf

this is the output (correct)

--2009-11-05 11:51:
08--  http://192.168.2.114:2480/base_c2.swf
Connecting to 192.168.2.114:
2480... connected.
HTTP request sent, awaiting response...
  HTTP/1.1
200 OK
  ETag: g1m02yog29nf
  Content-Type: application/x-shockwave-
flash
  Content-Length: 105819
  Date: Thu, 05 Nov 2009 11:51:08 GMT
 
Connection: Keep-Alive
Length: 105819 (103K) [application/x-shockwave-
flash]
Saving to: `base_c2.swf'

100%
[==============================================================================================>]
105,819     --.-K/s   in 0s

2009-11-05 11:51:08 (337 MB/s) - `base_c2.
swf' saved [105819/105819]


but, if i do the same command 2-3 times,
wget need to retry it 3 times before correctly get the base_c2.swf...

this is really strange because the pc is obviusly the same machine
(wget machine is the machine that host GWS) and there is noone
connected to that machine, so servicing this file should be
always a
success without any retry! also, as a side note, i simply execute wget
3-4 times (UP arrow enter) and they should't be concurrent afaik


3)

this is the most difficult to explain and reproduce...
internet
explorer 8 only bug :S
a bit of history on how the system work
i have 2
connection a <push over pull> and a <normal pull> request
so,
basically, on flash i have 2 different connector that work as follow

<PULL> ---> GWS ---> <RESPONSE>---> CLIENT
<PUSH> ---> GWS ---> WAIT
TILL EVENT---> <RESPONSE>---> CLIENT
this always worked like a charm,
on .16
this kind of work is handled via a asyncexecutorservices, btw.

unfortunatly with .18d what's happen is that
after a couple of refresh
of the browser, the socket doesn't work anymore in an indipendent way,
and every request seems to be serialized.
i had the idea that the issue
was on keepalive, but probably i was wrong.

what solved the problem,
was to put a lighttpd with a mod_proxy before GWS webserver :S

this
isn't really convenient.


If you need more help, let me know,
unfortunatly i can't provide any test case, due to the difficulty of
reproduce such kind of error :(
i have already try with a plain
webserver, but try to implement a push over pull and a pull for flash
is quite a long work :(

take care!









Passa a Tiscali Tutto Incluso Light: telefono + adsl 8 Mb senza limiti a soli 9,95 euro al mese fino al 01/04/2010. Gratis la Sim Tiscali Mobile con 25 euro di traffico.

L’offerta è valida solo se attivi entro il 05/11/09

http://abbonati.tiscali.it/telefono-adsl/prodotti/tc/tuttoincluso_light/?WT.mc_id=01fw

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@...
For additional commands, e-mail: dev-help@...

[Jeanfrancois Arcand-2]

Salut,

rama.rama@... wrote:

> Hi all,
>
>
> after an update to gws 1.9.18d, from 1.9.16 i have noticed
> some issue that i want to share with you
>
>
> 1)
> when i do a request for
> a certain page (that should return a big xml) i got
> <fancy xml>
>
> <blabla/>
> <blabla/>
> <blabla/>
> Http header (!?!?!)
> <blabla/>
> <blabla/>
>
> <blabla/>
> </fancy xml>
>
> this bug was already present in the past, and
> if i remember well was a problem with http1.0 or something like that (i
> don't know if this is till the case)
> In any case, what's happen is that
> the header are sended after a piece of request is issued :(

Ok so this is not a regression, right? But can it be your adapter that
doesn't flush the headers before writing bytes? Can you share you
Adapter privately? Push it to jfarcand at apache.org

>
>
> 2)
> wget
> issue
> if i do a
> prometeo:~# wget -S -v http://192.168.2.114:
> 2480/base_c2.swf
>
> this is the output (correct)
>
> --2009-11-05 11:51:
> 08--  http://192.168.2.114:2480/base_c2.swf
> Connecting to 192.168.2.114:
> 2480... connected.
> HTTP request sent, awaiting response...
>   HTTP/1.1
> 200 OK
>   ETag: g1m02yog29nf
>   Content-Type: application/x-shockwave-
> flash
>   Content-Length: 105819
>   Date: Thu, 05 Nov 2009 11:51:08 GMT
>  
> Connection: Keep-Alive
> Length: 105819 (103K) [application/x-shockwave-
> flash]
> Saving to: `base_c2.swf'
>
> 100%
> [==============================================================================================>]
> 105,819     --.-K/s   in 0s
>
> 2009-11-05 11:51:08 (337 MB/s) - `base_c2.
> swf' saved [105819/105819]
>
>
> but, if i do the same command 2-3 times,
> wget need to retry it 3 times before correctly get the base_c2.swf...

Is FileCaching enabled? Looks like a bug with FileCache *and* Grizzly
ARP. What is the size of the file?

>
> this is really strange because the pc is obviusly the same machine
> (wget machine is the machine that host GWS) and there is noone
> connected to that machine, so servicing this file should be
> always a
> success without any retry! also, as a side note, i simply execute wget
> 3-4 times (UP arrow enter) and they should't be concurrent afaik
>
>
> 3)
>
> this is the most difficult to explain and reproduce...
> internet
> explorer 8 only bug :S
> a bit of history on how the system work
> i have 2
> connection a <push over pull> and a <normal pull> request
> so,
> basically, on flash i have 2 different connector that work as follow
>
> <PULL> ---> GWS ---> <RESPONSE>---> CLIENT
> <PUSH> ---> GWS ---> WAIT
> TILL EVENT---> <RESPONSE>---> CLIENT
> this always worked like a charm,
> on .16
> this kind of work is handled via a asyncexecutorservices, btw.
>
> unfortunatly with .18d what's happen is that
> after a couple of refresh
> of the browser, the socket doesn't work anymore in an indipendent way,
> and every request seems to be serialized.
> i had the idea that the issue
> was on keepalive, but probably i was wrong.
>
> what solved the problem,
> was to put a lighttpd with a mod_proxy before GWS webserver :S
>
> this
> isn't really convenient.

Ok that one it will be difficult to reproduce, but there is clearly a
regression. I will go over the commit and see what's happening. Can you
grab a network dump between the client and the server?

Also, are you 100% sure IE 8 is the same version tested on 1.9.16 and
1.9.18?



>
>
> If you need more help, let me know,
> unfortunatly i can't provide any test case, due to the difficulty of
> reproduce such kind of error :(
> i have already try with a plain
> webserver, but try to implement a push over pull and a pull for flash
> is quite a long work :(
>

I understand. Please share as much info as you can privately so we can help.

A+

-- Jeanfrancois

> take care!
>
>
>
>
>
>
>
>
>
> Passa a Tiscali Tutto Incluso Light: telefono + adsl 8 Mb senza limiti a soli 9,95 euro al mese fino al 01/04/2010. Gratis la Sim Tiscali Mobile con 25 euro di traffico.
>
> L’offerta è valida solo se attivi entro il 05/11/09
>
> http://abbonati.tiscali.it/telefono-adsl/prodotti/tc/tuttoincluso_light/?WT.mc_id=01fw
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@...
> For additional commands, e-mail: dev-help@...
>

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@...
For additional commands, e-mail: dev-help@...