Changing... serial number back in the past.

Dan Mahoney, System Admin danm at prime.gushi.org
Thu Oct 5 21:45:25 UTC 2006


On Thu, 5 Oct 2006, Kevin Darcy wrote:

> aad wrote:
>> On 10/5/06, Dawn Connelly <dawn at zapata.org> wrote:
>>
>>> You'll need to delete the copy of the zone file on the slave servers. Otherwise
>>> they will not pull over the correct serial number.  Remove, reload, and you are
>>> all set.
>>>
>>
>> Or, if you don't want to have to stop the slave servers, you could
>> simply set the serial number to 0 on the master server and wait for
>> all the slaves to pick it up. Then, reset the serial on the master to
>> the desired value.
>>
> Bad advice. Unless the slaves see the master's serial number as "higher"
> than their own, they'll never transfer the zone. One could wait for that
> replication indefinitely.
>
> Read RFC 1982 to see how serial number arithmetic really works. A
> "reset" is possible, but setting the serial to 0 isn't always the way to
> accomplish it.

The method in rfc1982 is especially useful when you DO NOT have control of 
one or more of the slave servers.  I've found it to be excessive to have 
to wait for the zones to retransfer when you can do "plan b":

In cases where I do have control of ALL the slaves -- I usually update the 
master, with reload, then delete the zonefiles from the slaves, and then 
*completely stop* and restart bind.  Reloading, refreshing, HUP'ing, none 
of that is enough to make that slave server forget the serial number it's 
got cached, but if bind starts up with no zonefile, it will start fresh.

An important point most people don't get immediately: the serial number 
ONLY matters between masters and slaves -- it has nothing to do with when 
people querying you will refresh.  The only numbers in your zonefile that 
matter to anyone you don't have a master/slave relationship there is the 
TTL.

--

--------Dan Mahoney--------
Techie,  Sysadmin,  WebGeek
Gushi on efnet/undernet IRC
ICQ: 13735144   AIM: LarpGM
Site:  http://www.gushi.org
---------------------------



More information about the bind-users mailing list