Andrew Sullivan <ajs at crankycanuck.ca> writes:
> On Thu, Sep 18, 2008 at 02:55:17PM -0400, Knight, Doug wrote:
>> page of the Slony-I site, with no further explanation. Looking at the
>> FAQ, section 5.5 the bug is mentioned again, with a note that it "should
>> be resolved some time after Slony-I 1.1.0". So first, what exactly is
>> the bug, and second does this bug still exist in 1.2.14?
>
> I don't know if the bug was fixed (or if it was fixable without really
> bad performance effects), but the problem was that sequence-only sets
> wouldn't replicate at all.

It was stranger than that.

The problem was that if you had a series of subscription sets coming
from a provider, but had NO tables in that series, then the query that
pulls the list of tables to replicate would find none, and fail.

If you had *any* tables being replicated, across a series of sets,
then all would work fine; the problem was *only* when you had no
tables at all.

Thus, if you have:

  set 1, consisting of tables t1, t2, t3, and sequences s1, s2, and s3, and
  set 2, consisting of sequences s4, s5, and s6,

then everything would work fine if:
  a) Node 22 is subscribing to node 11, and pulling *BOTH* set 1 and set 2, or
  b) Node 22 is subscribing to node 11, and pulling just set 1.

The case that fails is where
  c) Node 22 is subscribing to node 11, and pulling just set 2.

Going over CVS logs, the issue of bug #1226 was addressed in version
1.86.2.4 of src/slon/remote_worker.c, on 2005-10-07 18:08:39.  Should
be long, long, long, gone.
-- 
(format nil "~S@~S" "cbbrowne" "cbbrowne.com")
http://linuxdatabases.info/info/internet.html
Rules of the  Evil  Overlord #171.  "I  will not  locate  a base  in a
volcano, cave, or  any other location  where it would be  ridiculously
easy to bypass security by rappelling down from above."
<http://www.eviloverlord.com/>