This is quite the thread. Maybe we can generate this enthusiasm for other aspects of the project as well.

Does anyone want to volunteer to test the code by using
the odbc in both the sofia module as the reg db and from javascript
in some random application and simulate the db going down using odbc against several popular db variants and finding the best code to make sure that nothing goes wrong? I am more than happy to accept the result of such a vigorous endeavor over just speculation about what's best.

I have to do most of the code you see here by myself so I have limited
time to devote to certain aspects so please let me know if anyone wants to adopt some tasks and help make the project better.



Anthony Minessale II

FreeSWITCH http://www.freeswitch.org/
ClueCon http://www.cluecon.com/

AIM: anthm
MSN:anthony_minessale@hotmail.com
GTALK/JABBER/PAYPAL:anthony.minessale@gmail.com
IRC: irc.freenode.net #freeswitch

FreeSWITCH Developer Conference
sip:888@conference.freeswitch.org
iax:guest@conference.freeswitch.org/888
googletalk:conf+888@conference.freeswitch.org
pstn:213-799-1400


----- Original Message ----
From: Łukasz Zwierko <lzwierko@gmail.com>
To: freeswitch-users@lists.freeswitch.org
Sent: Wednesday, February 20, 2008 8:51:51 AM
Subject: Re: [Freeswitch-users] Spidermonkey ODBC

Hi,

I think you guys have a wrong approach to this issue, as far as I'm concerned, a well written application should have clear distinction between connection layer and sql layer. What I mean is that if application wants to do a query to DB, than handling issues like dropped connection should not be mixed with inserts/selects.
So, for example, there should exist a separate layer  which is responsible for generating queries, and separate layer which actually makes the query, handles errors and optionally, repeats the query or insert (refer to java spring/hibernate - quite heavy java engines).
This way you could avoid any "select 1" or such like stuff every time you want to do a transaction to DB.
Still you feel that "select 1" is a better choice, but not all DB support it, you could just make an assumption, that in each DB there shall exist a table with just one column and row, and make a query to it (can call it KeepAliveTable).

Still, in my opinion the application should not test the connection each time it wants to make a transaction as the overhead will be huge.

Łukasz


2008/2/20, Anthony Minessale <anthmct@yahoo.com>:
Many databases break silently or the connection drops
without any way to know for sure w/o trying to use it.

I choose to ensure in the core that the data is connected.
executing "select 1" is a small price to pay to make sure
that the database is connected considering the importance
of persistent connectivity in the sip, jingle and call
limit databases as well as the javascript that rely on it.

All of those applications to make the safe
assumption that the core will keep the db connected and
to change that would be drastic behavior change and
destabilize the code on the eve of our release.
 
Therefore i guess we should focus on better heuristics
to detect firebird, preferably based on the driver name
and not by doing a 2nd sql stmt every time.


 
Anthony Minessale II

FreeSWITCH http://www.freeswitch.org/
ClueCon http://www.cluecon.com/

AIM: anthm
MSN:anthony_minessale@hotmail.com
GTALK/JABBER/PAYPAL:anthony.minessale@gmail.com
IRC: irc.freenode.net #freeswitch


----- Original Message ----
From: David Revill <davidrevill@datarun.co.uk>
To: freeswitch-users@lists.freeswitch.org
Sent: Tuesday, February 19, 2008 4:37:28 PM
Subject: Re: [Freeswitch-users] Spidermonkey ODBC

Unfortunately, there isn't a universal ansi sql stmt that will do it as the only statements which can be guaranteed to follow the ansi standard are the simple select, update and delete statements which rely on knowing some table name . 
 
I think we are trying to solve a problem which need not exist. The Javascript application should handle the idle timeout. It should close the connection if it will not be using it for a while, and,if it needs to, can do what db_is_up is attempting to do in the application code. Testing for connection before every query seems an unneccessary overhead.
 
----- Original Message -----
Sent: Tuesday, February 19, 2008 6:38 PM
Subject: Re: [Freeswitch-users] Spidermonkey ODBC

is there some universal ansi sql stmt that is small and harmless
but is supported by everything? to replace "select 1"


 
Anthony Minessale II

FreeSWITCH http://www.freeswitch.org/
ClueCon http://www.cluecon.com/

AIM: anthm
MSN:anthony_minessale@hotmail.com
GTALK/JABBER/PAYPAL:anthony.minessale@gmail.com
IRC: irc.freenode.net #freeswitch


----- Original Message ----
From: David Revill <davidrevill@datarun.co.uk>
To: freeswitch-users@lists.freeswitch.org
Sent: Tuesday, February 19, 2008 6:56:31 AM
Subject: Re: [Freeswitch-users] Spidermonkey ODBC

One way to detect firebird would be to put the 'Select first 1  * from RDB$relations' into switch_odbc_handle_connect. If it succeeds then it is Firebird. A better query might be 'select * from rdb$database' as the system table rdb$database only ever has 1 record.
----- Original Message -----
Sent: Tuesday, February 19, 2008 12:18 PM
Subject: Re: [Freeswitch-users] Spidermonkey ODBC

I''ve only tested it with firebird on windows. As you said, your driver name doesnt match any of the names I used to detect if it's firebird.

The problem is that most databases supports "SELECT 1" as a query. It will always be successful if the the connection is up. However, firebird did not support "SELECT 1" and therefore i did query the system database instead.

If the firebird check fails, it will try to invoke "SELECT 1", and on firebird that query will fail. And the failure is interpreted as the connection is down, and therefore it tries to reconnect.

That's probably why you get that error. If you have any other idea on how I can detect if it's a firebird db, please let me know and i'll fix the odbc code.

On Feb 19, 2008 1:05 PM, David Revill <davidrevill@datarun.co.uk> wrote:
That's a fair point, but db_is_up is only called in 2 places and the attempted re-connect could be done if the exec'd query fails. That way it would not be dependant on the name of the driver.
 
In my application, the database was queried at the beginning of the script and the connection closed. It was explicitly re-connected at the end to post data, so the connection timeout was not an issue.
----- Original Message -----
Sent: Tuesday, February 19, 2008 10:35 AM
Subject: Re: [Freeswitch-users] Spidermonkey ODBC

I did that code a couple of months ago.
The function was added since connections are closed if they are idle too long, and they were not reconnected again if that happened.


On Feb 19, 2008 11:22 AM, David Revill <davidrevill@datarun.co.uk> wrote:
The problem is in switch_odbc.c function db_is_up. It assumes that the variable is_firebird is true, but the setting of this variable assumes that the Firebird driver name has 'FIREBIRD', 'FB64' or 'FB32' in it.
 
I commented out  the code in db_is_up and just returned a true value, as on closer inspection could not see the point of this function.
 
David Revill
----- Original Message -----
From: David Revill
Sent: Tuesday, February 19, 2008 8:58 AM
Subject: Re: [Freeswitch-users] Spidermonkey ODBC

I spotted this error a few months ago, and to my shame did not let everyone know.  I have since archived the project so cannot point you to the actual code. However, the problem is  that it assumes the name of the Firebird Driver (FB64?)  to set a boolean, so that it handle the connection test differently.
 
When looking at the actual code, I actually felt that the connection test was redundant, as all it did was a select from a system table. This meant it was doing two queries for everyone required. (One to see if it could do a query, and then one to do it). I therefore commented the connection test function out.
 
regards
 
David Revill
----- Original Message -----
From: Steven Brown
Sent: Monday, February 18, 2008 11:34 PM
Subject: [Freeswitch-users] Spidermonkey ODBC

Hi,
 
I'm experimenting with spidermonkey for JavaScript call control, the basics seem fine and I'm now trying to connect to an existing Firebird database, the Firebird odbc lib is installed ok as is unixODBC and I can confirm this and access the db no problem with isql, something strange happens though when I connect in spidermonkey, basically the odbc connection is made successfully, but then as soon as I call either exec or query the odbc connection starts to drop and re-connect continuously,
 
the isql output is below
 
isql -v test
+---------------------------------------+
| Connected!                            |
|                                       |
| sql-statement                         |
| help [tablename]                      |
| quit                                  |
|                                       |
+---------------------------------------+
SQL> select first 1 * from pool_phones
+---------------------+------------+
| GSMNO               | ORGNO      |
+---------------------+------------+
| 0712345678         | 1          |
+---------------------+------------+
SQLRowCount returns 1
1 rows fetched
SQL>
however the following test code calling the same query
 
use("ODBC");
var db = new ODBC("test","SYSDBA","masterkey");
db.connect();
db.query("select first 1 *  from pool_phones");
db.nextRow();
row = db.getData();
console_log("INFO","HELLO " +  row["GSMNO"]  + "\n");
exit();
 
gives the following output and then just loops disconnecting and reconnecting until I shutdown freeswitch
2008-02-18 23:26:03 [DEBUG] switch_core_state_machine.c:144 switch_core_standard_on_execute() sofia/default/1000@192.168.0.7:5060 Execute javascript(/usr/scripts/test1.js)
2008-02-18 23:26:03 [DEBUG] mod_spidermonkey.c:3150 js_api_use() Loading ODBC
2008-02-18 23:26:03 [DEBUG] switch_odbc.c:145 switch_odbc_handle_connect() Connecting test
2008-02-18 23:26:03 [DEBUG] switch_odbc.c:174 switch_odbc_handle_connect() Connected to [test]
2008-02-18 23:26:03 [DEBUG] switch_odbc.c:95 switch_odbc_handle_disconnect() Disconnected 0 from [test]
2008-02-18 23:26:03 [DEBUG] switch_odbc.c:142 switch_odbc_handle_connect() Re-connecting test
2008-02-18 23:26:03 [DEBUG] switch_odbc.c:145 switch_odbc_handle_connect() Connecting test
2008-02-18 23:26:03 [DEBUG] switch_odbc.c:174 switch_odbc_handle_connect() Connected to [test]
2008-02-18 23:26:03 [CRIT] switch_odbc.c:234 db_is_up() The sql server is not responding for DSN test []
2008-02-18 23:26:03 [INFO] switch_odbc.c:239 db_is_up() The connection has been re-established
...

Any ideas much appreciated

Thanks

Steve

 

 

 

 



 
 
 
 

_______________________________________________
Freeswitch-users mailing list
Freeswitch-users@lists.freeswitch.org
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org

_______________________________________________
Freeswitch-users mailing list
Freeswitch-users@lists.freeswitch.org
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org

_______________________________________________
Freeswitch-users mailing list
Freeswitch-users@lists.freeswitch.org
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org



_______________________________________________
Freeswitch-users mailing list
Freeswitch-users@lists.freeswitch.org
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org


_______________________________________________
Freeswitch-users mailing list
Freeswitch-users@lists.freeswitch.org
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org


-----Inline Attachment Follows-----

_______________________________________________
Freeswitch-users mailing list
Freeswitch-users@lists.freeswitch.org
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org


Be a better friend, newshound, and know-it-all with Yahoo! Mobile. Try it now.

_______________________________________________
Freeswitch-users mailing list
Freeswitch-users@lists.freeswitch.org
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org


-----Inline Attachment Follows-----

_______________________________________________
Freeswitch-users mailing list
Freeswitch-users@lists.freeswitch.org
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org


Looking for last minute shopping deals? Find them fast with Yahoo! Search.

_______________________________________________
Freeswitch-users mailing list
Freeswitch-users@lists.freeswitch.org
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org




-----Inline Attachment Follows-----

_______________________________________________
Freeswitch-users mailing list
Freeswitch-users@lists.freeswitch.org
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org


Looking for last minute shopping deals? Find them fast with Yahoo! Search.