[asterisk-dev] The New CDR system

[Steve Totaro]

Leif Madsen wrote:
> On Saturday 31 March 2007 17:37:30 zoachien at securax.org wrote:
>   
>> Steve Totaro wrote:
>>     
>>> Vasil Kolev wrote:
>>>       
>>>> A question that nobody seems to have touched yet on this (and maybe the
>>>> reason is that this is not the right discussion :) ), what about a way
>>>> to specify a backup storage for the CDR data? For example your DB dies,
>>>> and suddenly you can't log it - how does it sound to have another way to
>>>> keep the data and to send it to the primary storage when it becomes
>>>> available?
>>>> (let's face it, everything crashes once in a while, no matter how
>>>> foolproof and redundant it's made)
>>>>         
>>> Isn't that what DB clustering (Single Quorum) and RAID are for?
>>> I suppose a default to write to a rotating log file like queue_log as
>>> well as DB clustering would be even safe.
>>>       
>
> It doesn't need to be very complex, and most of this probably is handled by 
> having a multi-master replicated database setup (with PostgreSQL, this 
> basically means Slony-II or pgcluster).
>
> The only thing that would really need to be added would be the ability to 
> say, "hey, write to this database, but if its not available, write to the 
> secondary one". This is done in func_odbc by having more than one writehandle 
> which references two separate connections in res_odbc.
>
> Other than that, the rest of the "live backup" stuff is handled by the 
> database backend -- not by Asterisk as Steve was alluding to.
>
> Leif Madsen.
>   
Leif,

I was alluding to no connectivity to the database whatsoever in which 
case, Asterisk would write to a rotating log file to be parsed later.  
It is kind of difficult for the database backend to handle anything if 
it not reachable.

Thanks,
Steve