On Mon, 2006-10-02 at 12:29 -0400, Andrew Sullivan wrote:
> On Fri, Sep 29, 2006 at 11:41:38AM -0700, Jeff Davis wrote:
> > If I lock the set, then move the set, then I think it's already
> 
> I don't think you need to issue an explicit lock before the move.
> 

(in the command below, cluster_info.slonik only contains cluster name,
and node 1 & 2 admin conninfo)

% slonik
include <cluster_info.slonik>;
move set (id = 1, old origin = 1, new origin = 2);
<stdin>:2: PGRES_FATAL_ERROR select "_cluster01".moveSet(1, 2);  -
ERROR:  Slony-I: set 1 is not locked

But, if I do:

% slonik
include <cluster_info.slonik>;
lock set (id = 1, origin = 1);
move set (id = 1, old origin = 1, new origin = 2);

it works.

However, afterward, if I try to unlock, I get:

% slonik
include <cluster_info.slonik>;
unlock set (id = 1, origin = 1);
<stdin>:2: PGRES_FATAL_ERROR select "_cluster01".unlockSet(1);  - ERROR:
Slony-I: set 1 does not originate on local node

or

% slonik
include <cluster_info.slonik>;
unlock set (id = 1, origin = 2);
<stdin>:2: PGRES_FATAL_ERROR select "_cluster01".unlockSet(1);  - ERROR:
Slony-I: set 1 is not locked

which indicates to me that the "move set" also unlocked the set. So
what's the use of unlock set?

> > Under what circumstances should I use move set?
> 
> When you're hoping to move the origin.
> 

Oops, what I meant to say was, when should I use unlock set?

Thanks,
	Jeff Davis