We are running freeradius 0.9.3.1 on RH ES3. CDR accounting records from a Cisco AS5350 are logged to both a detail file and to Postgres SQL running on the same box. The issue appears to be the following:
For some calls, our PRI will terminate the call immediately because of unknown number, busy line, etc. So immediate, that freeradius receives both the start, start update, and stop records at basically the same time. The problem this creates is that it appears the insertion of the start record has not completed when the update for the start and then the stop record occurs (multiple handles to the database). This causes the update and stop records to “fall-thru” the update process and do an insertion of a full record for both. Thus I have instances of one CDR record that has three entries, (2 partial and 1 full) in SQL instead of the single entry that 99% of the other CDR record do. I haven’t decided if I should approach this from the Cisco side or from the freeradius side in the form of some type of delay or retry for SQL accounting records. I haven’t been able to find a freeradius configuration parameter that does this. Any ideas? I can provide more info if needed. - List info/subscribe/unsubscribe? See http://www.freeradius.org/list/users.html
