Le jeudi 30 juillet 2009 à 22:53 +0100, Dean Smith a écrit :
During a recent network incident we had some very high churn in PPP clients. Our NASes (Multiple Cisco 7200 as VPDN LNS) didn’t appear to be able to authenticate as quickly as they needed to. In the Radius packet logs we see quite a few Rejects for requests that should have been OK – indeed the identical requests were repeated and answered correctly seconds afterwards.
In the radius log itself we have quite a few of these:-
Mon Jul 27 22:51:24 2009 : Info: rlm_sql (sql): There are no DB handles to use! skipped 0, tried to connect 0 Mon Jul 27 22:51:25 2009 : Info: rlm_sql (sql): There are no DB handles to use! skipped 0, tried to connect 0 Mon Jul 27 22:51:29 2009 : Info: rlm_sql (sql): There are no DB handles to use! skipped 0, tried to connect 0
From the archives the usual reason for the above would seem to be the MySQL database not answering quickly enough. The relevant freeradius configs are at the defaults – but the box(s) themselves don’t appear to have struggled with CPU.
CPU is not the only bottle neck. For SGDB you should take care at IO more than cpu i think?
Would getting that error in the logs result in valid authentication requests being rejected ?
yes
Unfortunately I’m not very experienced in MySQL to be able to tune it. Would increasing the number of “num_sql_socks” within the sql.conf from the default 5 be a good starting point ? Are there other FreeRadius configs I can adjust before moving onto the MySQL database itself ?
You could start here. I personnaly have to open 40 sockets to my oracle database when freeradius starts.
Thanks
Dean
- List info/subscribe/unsubscribe? See http://www.freeradius.org/list/users.html