[Klone-users] Klone using server push
Stefano Barbato
barbato at koanlogic.com
Tue Jul 29 08:48:42 EDT 2008
Hi Andre,
KLone doesn't really know when a session (session_t object) expires;
every time a client comes KLone looks for its session-id in a cookie
and, if the given ID is valid and the expiration has not passed, it
loads the data on the server associated to that ID. It a client never
comes back there's no trigger the you can hook to be notified about
the expiration of its session data.
Note that session term() function is called on every request that use
the session object. Before running your own code in .kl1 pages KLone
create a few objects for you (request, response, session) and clean
them up and the end of custom code execution:
---------------
create request, response, session...
your_code
free request, response, session
---------------
If I understood correctely your question that is not what you need. An
easy solution would be adding a "last-accessed" field in your
structure that hold shared memory info and run a periodic cleanup
function that removes expired slots.
bye,
s
On 28/lug/08, at 17:56, Andre Puschmann wrote:
> Hi Stefano,
>
> once again, sorry for the delay :-)
>
> I am writing a klone-powered Linux packet capture application using
> libpcap, which updates a web-interface as soon as a new network packet
> arrives, i.e. live trace.
>
> I use a shared memory concept to access shared data within different
> processes. Furthermore, I use sessions to map a certain client
> connection to a certain part of the shared memory. However, after
> a session expires I need to do some memory cleanups in the shared
> area, i.e. free reserved slots for further client connection.
>
> I am not sure about the desired way to do so. The only thing I could
> find was a private term function in each session object.
>
> Best regards,
> Andre
>
>
>
> Stefano Barbato wrote:
>> Hi Andre,
>>
>> the method to use really depends on what you need to accomplish.
>>
>> If you need to share per-client data you can use sessions (url [1]).
>>
>> If you want to share data between server processes you can use any
>> facility the operating system offer (which one do you use btw?) like
>> shared memory objects ([2]), memory mapped file [3], external file-
>> based
>> database like sqlite [4], etc.
>>
>> If you want more then bare data sharing you can use an external
>> daemon
>> that centralizes klone daemons calls and that accept requests on
>> tcp/ip
>> or unix sockets. Even a shell script called by inetd may be enough
>> for
>> example. It really depends on your requiremets.
>>
>> The "iterative" model you tried allow just one connection at a time
>> so
>> may not be suitable for your particular need. If you need multiple
>> concurrent access to the same KLone application you need to use the
>> "fork" or "prefork" model type.
>>
>> So there are many options but I need more info to point you to the
>> right
>> direction :)
>>
>> bye,
>> s
>>
>>
>> [1] sessions: http://www.koanlogic.com/klone/api/html/group__session.html
>> [2] man shm_open
>> [3] mmap
>> [4] www.sqlite.org
>>
>>
>> On 16/lug/08, at 23:10, Andre Puschmann wrote:
>>
>>> Hi Stefano,
>>>
>>> thanks for your quick reply.
>>> It worked as you described, thank you very much. However, I now got
>>> another problem/question.
>>>
>>> I've read that klone runs multiple processes to serve client
>>> connections
>>> (prefork pattern).
>>>
>>> Is it possible to have shared resources for all processes?
>>>
>>> Lets say I want to have just _one_ socket descriptor (singleton)
>>> that
>>> I want to use.
>>> I tried to use the iterative model to do so. But if that single
>>> descriptor does a blocking read, the server hangs. So this is no
>>> solution. Would you recommend using pthread for this?
>>>
>>> I am not sure if I get something wrong :-)
>>>
>>> Regards,
>>> Andre
>>>
>>>
>>>
>>>
>>>
>>>
>>> Stefano Barbato wrote:
>>>> Hi Andre,
>>>>
>>>> KLone buffers the first 4096 bytes to give users the opportunity to
>>>> modify header fields even after a few bytes have been written out
>>>> with
>>>> io_printf or io_write. You can for example use the
>>>> response_redirect
>>>> even after io_prints:
>>>>
>>>> <%
>>>> io_printf(out, "result 1");
>>>> io_printf(out, "result 2");
>>>> io_printf(out, "result 3");
>>>>
>>>> /* then if an error occurred you can still modify http header */
>>>> if(error_condition)
>>>> response_redirect(response, "/error.kl1");
>>>> %>
>>>>
>>>> The buffer is implemented as a filter attached to the response_t
>>>> I/O
>>>> object so you can avoid such buffer removing such filter and
>>>> explicitly
>>>> call the response_print_header() function (if you want an HTTP
>>>> header at
>>>> least). Here's the code:
>>>>
>>>>
>>>> <%
>>>> ...
>>>> response_set_field(response, "Expires:","Mon, 26 Jul 1997
>>>> 05:00:00
>>>> GMT");
>>>>
>>>> io_codecs_remove(response_io(rs));
>>>> response_print_header(response);
>>>>
>>>> io_printf(out, "---ThisRandomString---\n");
>>>> ...
>>>> %>
>>>>
>>>> Take a look to the "Message Channel" KLone demo application that
>>>> opens
>>>> and keep opened a bidirectional channel between client and server:
>>>>
>>>> http://koanlogic.com/download/klapp-msgchan-1.2.2.tar.gz
>>>>
>>>> It's built on Flash XmlSocket component and a custom easy library
>>>> we
>>>> wrote (see msgchan.h).
>>>>
>>>> bye!
>>>>
>>>> s
>>>>
>>>>
>>>>
>>>>
>>>> On 08/lug/08, at 19:32, Andre Puschmann wrote:
>>>>
>>>>> Hi all,
>>>>>
>>>>> I am trying to use a server push technique to avoid client side
>>>>> polling
>>>>> for data.
>>>>> I am not sure if I am doing something wrong. For some reason Klone
>>>>> sends
>>>>> all 4 "packets" in one single bunch. Well, at least wireshark
>>>>> reports
>>>>> this!
>>>>> Seems like io_flush doesn't work?!
>>>>>
>>>>> Here is my sample code:
>>>>>
>>>>> <code>
>>>>> <%!
>>>>> #include <stdio.h>
>>>>> %><%
>>>>> int i;
>>>>>
>>>>> response_set_content_type(response,
>>>>> "multipart/x-mixed-replace;boundary=---ThisRandomString---");
>>>>> response_set_field(response, "Pragma:", "no-cache");
>>>>> response_set_field(response, "Cache-Control", "no-cache, no-store,
>>>>> must-revalidate");
>>>>> response_set_field(response, "Expires:", "Mon, 26 Jul 1997
>>>>> 05:00:00
>>>>> GMT");
>>>>>
>>>>>
>>>>> io_printf(out, "---ThisRandomString---\n");
>>>>>
>>>>> for (i = 1; i < 5; i++)
>>>>> {
>>>>> io_printf(out, "Content-type: text/html\n\n");
>>>>> io_printf(out, "<h2>part %d</h2>\n", i);
>>>>> io_printf(out, "---ThisRandomString---\n");
>>>>> io_flush(out);
>>>>> sleep(1);
>>>>> }
>>>>> %>
>>>>> </code>
>>>>>
>>>>> Using the built-in cgi works for a shell script. Also here's the
>>>>> code
>>>>> for that one:
>>>>>
>>>>> <code>
>>>>> #!/bin/sh
>>>>> echo "HTTP/1.0 200"
>>>>> echo "Content-type:
>>>>> multipart/x-mixed-replace;boundary=---ThisRandomString---"
>>>>> echo ""
>>>>> echo "---ThisRandomString---"
>>>>> while true
>>>>> do
>>>>> echo "Content-type: text/html"
>>>>> echo ""
>>>>> echo "time: "
>>>>> date
>>>>> echo "---ThisRandomString---"
>>>>> sleep 5
>>>>> done
>>>>> </code>
>>>>>
>>>>>
>>>>> Comments are welcome!
>>>>>
>>>>> Best regards,
>>>>> Andre
>>>>>
>>>>> _______________________________________________
>>>>> Klone-users mailing list
>>>>> Klone-users at koanlogic.com
>>>>> http://koanlogic.com/cgi-bin/mailman/listinfo/klone-users
>>>>>
>>>>
>>>>
>>>> _______________________________________________
>>>> Klone-users mailing list
>>>> Klone-users at koanlogic.com
>>>> http://koanlogic.com/cgi-bin/mailman/listinfo/klone-users
>>>>
>>>>
>>>
>>>
>>> _______________________________________________
>>> Klone-users mailing list
>>> Klone-users at koanlogic.com
>>> http://koanlogic.com/cgi-bin/mailman/listinfo/klone-users
>>>
>>
>>
>> _______________________________________________
>> Klone-users mailing list
>> Klone-users at koanlogic.com
>> http://koanlogic.com/cgi-bin/mailman/listinfo/klone-users
>>
>>
>
>
>
>
> _______________________________________________
> Klone-users mailing list
> Klone-users at koanlogic.com
> http://koanlogic.com/cgi-bin/mailman/listinfo/klone-users
>
More information about the Klone-users
mailing list