On 24 Jul 2012, at 10:34, Arran Cudbard-Bell wrote:

> The accounting section of rlm_sql has been modified to use reference based 
> accounting queries.
> 
> Accounting and post-auth now have their own subsections, and contain a config 
> pair called reference. The value of this is expanded to give a config path, 
> and the config pair this resolves to is used as the query value.
> 
> If the query fails and the config path matches multiple pairs the next pair 
> is used. This emulates the 'alt' behaviour that was were before but can fail 
> between N queries. 
> 
> I've tested the MySQL version pretty thoroughly, but it'd be nice if people 
> using Postgresql, Mssql, and Oracle could also test, just to make sure no 
> query errors have crept in with the reformatting.
> 
> Code is available here: 
> https://github.com/arr2036/freeradius-server/tree/dynamic_acct_query_selection
> 
> There's also a patch in that branch from Alan to strip off whitespace from 
> the start ofline continuations if the config pair value begins with "\\n - 
> slightly magic but seems to be pretty effective - debug output is much more 
> readable.
> 
> The MySQL schemas have also been updated with new features. If anyone fancies 
> adding the following to the default schemas/configs it'd be much appreciated:
>       * Calculating all times from %{integer:Event-Timestamp} (you'll have to 
> convert from unix timestamp using an SQL function) and removing the 
> acctstartdelay and acctstopdelay columns.
>       * Adding an acctupdatetime column, and updating it when a new session 
> is created or an interim-update packet is received.
>       * Adding an acctinterval column and updating it using the delta (in 
> seconds) between the last acctupdatetime value and the current time (this is 
> really easy way to do stale session detection if your NAS supports 
> interim-updates). This is difficult in some SQL dialects, because there's no 
> guarantee that fields will be modified in the order specified in the query.
>       * Removing the xascend column, and using a single list of fields for 
> all insert operations.
> 
> Look at the MySQL schema if the above are unclear, and please post to the 
> devel list if you want to work on one of the schemas/databases, so we don't 
> have duplicate effort.

And if this guy is still on the list… 

http://www.mail-archive.com/[email protected]/msg42306.html

It's now supported, only took four years :)

-Arran

-
List info/subscribe/unsubscribe? See http://www.freeradius.org/list/users.html

Reply via email to