Re: Bind918 slave reports Dumping master file...open: file not found

> On Feb 17, 2024, at 9:39 PM, Tatsuki Makino <tatsuki_makino@hotmail.com> wrote:
> 
> Hello.
> 
> It would be easier for later visitors to understand if the background of the "working" directory were also explained.
> 
> Moin Rahman wrote on 2024/02/18 03:34:
>>> On Feb 17, 2024, at 7:26 PM, bob prohaska <fbsd@www.zefox.net> wrote:
>>> The options section of named.conf contains:
>>>       directory       "/usr/local/etc/namedb/working";
>> 
>> Do you have a directory named `slave` under `/usr/local/etc/namedb/working` ?
> 
> First, this directory is just to avoid the warning that occurs at
> https://gitlab.isc.org/isc-projects/bind9/-/blob/366b93f83565902aa98b3e146e0ead88e7502d87/bin/named/server.c#L9228
> 
> For this we will be asked to write a configuration file that takes into account that the working directory is there.
> In named.conf.sample, everything is replaced with absolute paths.
> 
> Since absolute paths are long, we want to use relative paths to make things easier :)
> So since all the provided directories are at the same level as the "working" directory, they need to be written with relative paths such as ../slave. and/or ../secondary..
> 
> Is this an acceptable explanation? :)
> 
> Regards.
> 

Technically yes. Historically not. :D


In some earlier versions the sample config we used to ship had relative
file paths I believe. And as I know that Bob has been a long term user
he indeed did not update the config file to the newer format where we
are using absolute filenames despite having directory directive.

So unless you know both the technical implications which you have
explained here and the historical implication it's difficult to
understand where these directories should be created.

But then again I haven't used bind for a long time so cannot say much.

Kind regards,
Moin