cyril.sarramiac at bt.com wrote:

>Hi,
>
>I am using Slony to replicate my database on a single slave database. It seems to work fine, but I have done several shell scripts to implement the failover and switch functionnalities. These scripts use slonik commands.
>Looking in Slony installation files, I found /etc/slon_tools.conf file which is used by perl scripts (under /usr/bin/, such as slon_kill). These scripts may be used to execute slony commands. I guess I could use these files for implementing my failover and switch functions in an other way than shell scripts.
>
>If I am right and before choosing my way of implementation, I would like to know if these scripts will be kept up to date in future versions, or, eventually, if one of the two ways (slonik commands or PERL scripts) will be given up. Can anybody answer?
>  
>
Well, there are a few considerations...

1.  Most of the Perl scripts are intended to generate slonik scripts
that you should ideally look at before running them.

2.  Maintenance of these scripts has been a bit "less than totally
tidy."  I prototyped them; someone else continued with them, but isn't
presently working on them.

3.  The regression tests don't make use of the scripts, so they are not
tested with anywhere near the frequency the rest of the system is tested.

4.  I foresee needing to make some significant revisions to the slon
management scripts in view of the 1.1 "log shipping" functionality.

Pointedly, I want to have the scripts use configuration files for each
slon so that we cut down on the risk of possibly forgetting to include
the "-a" option on log shipping nodes.  If you somehow restart the slon
without that option, this essentially "corrupts" the sequence of logs in
that some logs would not be written to disk.

At this point, I'd have to say that the Perl scripts represent
"hopefully useful samples" as opposed to being a permanently
authoritative "API" of sorts.