forcing zone file update with DDNS

[Jeff Layton]

Ahh, I did check the source code, but hadnt found it yet (this was my first
perusal of the BIND source). Thanks for the info, I feel more mollified knowing
that bind can roll back the .log files to do the updates :-).
-- Jeff

Jim Reid wrote:
> 
> >>>>> "Jeff" == Jeff Layton <jtlayton at pliantsystems.com> writes:
> 
>     >> It does periodically write out the zone.
> 
>     Jeff> How often is periodically (roughly)? Are we talking every 30
>     Jeff> mins or once a day?
> 
> Did you try reading the source code? From ns_defs.h:
> 
> #ifdef BIND_UPDATE
> #define SOAINCRINTVL 300 /* default value for the time after which
>                           * the zone serial number must be incremented
>                           * after a successful update has occurred */
> #define DUMPINTVL 3600   /* default interval at which to dump changed zones
>                           * randomized, not exact */
> #define DEFERUPDCNT     100     /* default number of updates that can happen
>                                  * before the zone serial number will be
>                                  * incremented */
> #define UPDATE_TIMER XFER_TIMER
> #endif /* BIND_UPDATE */
> 
>     Jeff> I'd hate to do a buncha updates and then lose them...  -- JL
> 
> That's why the name server writes out new updates to a .log
> transaction log file. It contains the dynamic updates which have been
> made to the zone since the last time a new zone file was written
> out. This allows the name server to roll forward by merging the old
> zone file with the transaction log to get the zone back to the state
> it was in before the name server crashed or whatever. So after
> approx. 1 hour, a new zone file is written out and the transaction log
> zapped because its updates have been incorporated into the new zone
> file. This means you should only lose updates if the filesystem loses
> files.

-- 
Jeff Layton					    jtlayton at pliantsystems.com