Askemos Forum

> First shot of a PostgreSQL driver, MySQL driver modified.
> 
> Hi all,
> 
> I just uploaded a first build of the BALL with PostgreSQL support.
> 
> Changes:
> 
> I) The formal of the XSQL connection definitions (distributed in
>    policy/xsql-definition.xml) has changed.  A connection has now a
>    'driver' child element (which might be moved in further releases to
>    become an attribute of the connection, I'm not sure what's the
>    right way here).  The content of that element define, which data
>    base driver is to be used.  Example
> 
>    <!-- This connection is universally open. -->
>    <connection>
>     <name>demo</name>
>     <driver>MySQL</driver>
>     <database>test</database>
>     <host>localhost</host>
>     <user>test</user>
>     <password></password>
>    </connection>
> 
>   Currently two values are supported "MySQL" and "PostgreSQL".
> 
> II) The mysql driver has changed slightly.  It used to block the whole
>    ball until the query completes.  At least Frank has suffered from
>    mysql queries taking so long, that BALL's heartbeat decided the
>    Askems server was locked up and killed it[1].  Unfortunately I've
>    not found a mysql API, which allows asynchronous notification or
>    polling about query completion.  Therefore I just inserted a
>    configurable delay between the mysql query and response retrieval.
>    Call (set-mysql-query-time! (milliseconds <integer>)) in your
>    configuration to adjust that delay (defaults to 20 ms).
> 
>    Having written that, it turned out, that mysql_real_query is to
>    blame.  It turned out that the ball stoped during this call.  Does
>    someone have some info how to tune mysql not to block there?

hi Jerry,

after updating rscheme and askemos now I test some applications local.
I change the connection parameters how descripted in I.

This error message occured by any mysql requests ?

Sat, 06 Dec 2003 23:51:19 +0100 <simple-error> error: sql-field: generic function doesn't understand #[<mysql-result> @087d_56fb]
(rest are (0))
 ((stack . @085b_393b))


So I testing the new tuning parameter in config - askemos silence, how you noted 

heartbeat: "Kernel did not send alive signal, killing it."


I overload some more in the config, grapped from mysql.scm

(define *mysql-query-time* 1000)
(define (set-mysql-query-time! (n <integer>))
  (set! *mysql-query-time* n))

now askemos isn't hanging, but may be ignored

Our applications based on the mysql db, so I must stop the test on this point and waiting.
  
bye Frank

 


> 
> III) The new PostgreSQL driver is almost identical to the mysql
>    driver, except that PostgreSQL has 'PQisBusy', which allows me to
>    poll for query completion.  Though this is better (and should
>    therefore become the recommented way to use XSQL within BALL) than
>    the mysql interface, it's still not satisfactory.  If anyone could
>    point me to some notification interface, possibly with example
>    code, that would be great.
> 
>    For the time being, you can adjust the polling interval in your
>    configuration using (set-pgsql-query-time! (milliseconds <integer>))
> 
> Please note that this code is not yet tested and I will not have the
> time to test it right now.  Please help.
> 
> Best regards
> 
> /J�rg
> 
> Footnotes:
> 
> 1) Actually I'm curious to see what huge or complicated data base
>    query will keep mysql busy for more than two seconds.  Frank, can
>    you provide some details?
> 
> -- 
> The worst of harm may often result from the best of intentions.