Christopher Browne cbbrowne at ca.afilias.info
Wed Mar 7 15:32:41 PST 2007
We are pleased to release versions 1.1.7 and 1.2.7, which contain
fixes for various problems found in these branches over the last
several months.

Below are listed the release notes for these two releases.  Please see
<http://main.slony.info/> to find CVS and downloadable tarballs.

----------------------------------------------------------------------
$Id: RELEASE-1.1.7,v 1.1.2.6 2007-03-06 18:47:23 cbbrowne Exp $

- Add remote_listen_timeout parameter to slon.conf

This addresses the problem where a slon times out when accessing
sl_event if a node has been out of commission for a long time (several
days)

- Add node numbers to error reports in slonik

If portions of UNINSTALL NODE break, report in the error message which node
it was working on.  That way the gentle user gets an indication as to 
which node 'broke' if the slonik script contained multiple such
requests.

- Applied bug fix for #1538

If cluster has only one node, then remove all events up to the last
SYNC.  That allows the cleanup loop to clear out sl_log_{1/2}.
Otherwise, the log tables will forever bloat until you add a second
node...

- Added test to test1 for function generate_sync_event() and make_function_strict

- Added "v81" files (for slony1_base.v81.sql, slony1_funcs.v81.sql,
xxid.v81.sql), necessary to support 8.1 "ALTER FUNCTION ... STRICT";

- Fixed quoting problem in generate_sync_event()

- Added functionality to UPDATE FUNCTIONS to make xxidin() function
STRICT; the absence of this caused postmaster to fall over when
processing MOVE SET event in PG v8.2 (not to say that Slony-I 1.1 now
*supports* 8.2; it does not)

- When you run MOVE SET, this populates sl_setsync for the moved set
  even on nodes that are not subscribed.  If, subsequent to doing this, 
  you attempt a SUBSCRIBE SET for a formerly-unsubscribed node, the
  subscription will fail right at the end when the slon tries to insert a
  new value to sl_setsync.

  The fix: DELETE from sl_setsync immediately before the INSERT.  This
  will silently blow away any 'offending' sl_setsync row.

  (As observed by Afilias staff...)

- Log shipping fix - storage of sl_setsync_offline call had a wrong
  printf type; change from %d to %s

----------------------------------------------------------------------
$Id: RELEASE-1.2.7,v 1.1.2.7 2007-03-06 18:47:45 cbbrowne Exp $

- Add remote_listen_timeout parameter to slon.conf

This addresses the problem where a slon times out when accessing
sl_event if a node has been out of commission for a long time (several
days)

- Resolve bug #1623

In this bug, big "action lists" that need to get compressed could cause
a logging printf to blow up.  Changed the logging level so that detail
is only shown at level 4, which won't bite people by default.

- UNINSTALL NODE failures now show node # in slonik error messages

If a user ran several UNINSTALL NODE requests in a single slonik
script, and one of them broke, you'd have no ready way to tell which
node this failed on.  Added code to report the node # where it failed.

- Added test to test1 for function generate_sync_event() and make_function_strict

- Added "v81" files (for slony1_base.v81.sql, slony1_funcs.v81.sql,
xxid.v81.sql), necessary to support 8.1 "ALTER FUNCTION ... STRICT";

- Fixed quoting problem in generate_sync_event()

- Added functionality to UPDATE FUNCTIONS to make xxidin() function
STRICT; the absence of this caused postmaster to fall over when
processing MOVE SET event in PG v8.2

- Added documentation of an issue surrounding NULLABLE columns to the
log analysis chapter of the admin guide and to the UPGRADING docs.

- When you run MOVE SET, this populates sl_setsync for the moved set
  even on nodes that are not subscribed.  If, subsequent to doing this, 
  you attempt a SUBSCRIBE SET for a formerly-unsubscribed node, the
  subscription will fail right at the end when the slon tries to insert a
  new value to sl_setsync.

  The fix: DELETE from sl_setsync immediately before the INSERT.  This
  will silently blow away any 'offending' sl_setsync row.

  (As observed by Afilias staff...)

- Log shipping fix - storage of sl_setsync_offline call had a wrong
  printf type; change from %d to %s
-- 
(reverse (concatenate 'string "ofni.sailifa.ac" "@" "enworbbc"))
<http://dba2.int.libertyrms.com/>
Christopher Browne
(416) 673-4124 (land)


More information about the Slony1-general mailing list