Christopher Browne [Thu, 28 Feb 2008 19:23:01 +0000 (19:23 +0000)]
Fixes needed to get "testpartition" to work on HEAD. The issue was that
with the new handling of triggers, it is no longer appropriate to run
the "restore table" function, as that function was only apropos for 1.2
and earlier, and has become deprecated in HEAD.
Christopher Browne [Wed, 27 Feb 2008 19:37:03 +0000 (19:37 +0000)]
Modifications to address things noted as unclear this week on the
Slony-I list:
1. Why is FORWARDING=yes typically needed on a subscriber node?
[because otherwise, you can't fail over to that node!]
2. RESTART NODE is a bit too strongly worded in claiming it is
useless after Slony-I 1.0.5.
Christopher Browne [Mon, 25 Feb 2008 17:42:19 +0000 (17:42 +0000)]
Per bug #37, add in a WAIT FOR EVENT in between the LOCK SET and
MOVE SET requests.
http://www.slony.info/bugzilla/show_bug.cgi?id=37
Christopher Browne [Mon, 25 Feb 2008 15:43:38 +0000 (15:43 +0000)]
Remove call to SP logswitch_weekly(), as we are now doing log switches
rather more frequently
Christopher Browne [Mon, 25 Feb 2008 15:42:32 +0000 (15:42 +0000)]
remove function logswitch_weekly() as we are now switching logs more or
less as often as possible
Christopher Browne [Mon, 25 Feb 2008 15:40:07 +0000 (15:40 +0000)]
Detect 3-argument version of typenameTypeId()
Christopher Browne [Mon, 25 Feb 2008 15:37:58 +0000 (15:37 +0000)]
Changed over to using entity references to point to sl_log_[12], sl_seqlog
Christopher Browne [Mon, 25 Feb 2008 15:31:25 +0000 (15:31 +0000)]
Elaborate on SYNC parameters to show a bit better how they interact with one
another
Christopher Browne [Thu, 14 Feb 2008 22:35:58 +0000 (22:35 +0000)]
Per Bug #36, -lpgport is only germaine to win/win32 platforms
Christopher Browne [Thu, 14 Feb 2008 22:21:42 +0000 (22:21 +0000)]
Expunge "SPOOLNODE" attribute that we had thought we would need for
log shipping. It expresses itself several ways:
1. sl_node.no_spool (in tables)
2. SPOOLNODE (in slonik grammars)
3. In various places, as an extra function argument
Christopher Browne [Thu, 14 Feb 2008 22:19:16 +0000 (22:19 +0000)]
Added real descriptions for options that were undocumented
Christopher Browne [Thu, 14 Feb 2008 16:41:35 +0000 (16:41 +0000)]
Apply changes for bug #31...
http://www.slony.info/bugzilla/show_bug.cgi?id=31
changes over to use PERLSHAREDIR in all of the scripts, and substitutes
that consistently across the various .pl files
Christopher Browne [Thu, 14 Feb 2008 16:38:08 +0000 (16:38 +0000)]
Mikhail Kolesnik pointed out some typos in slonik documentation.
Christopher Browne [Wed, 13 Feb 2008 23:02:40 +0000 (23:02 +0000)]
Replace some lingering references to old xxid functions to use the new
8.3 built-ins.
Christopher Browne [Wed, 13 Feb 2008 18:40:29 +0000 (18:40 +0000)]
change conditional in logswitch function to fold it into IF statement, and
eliminate the need to hide a SELECT.
Christopher Browne [Wed, 13 Feb 2008 18:39:17 +0000 (18:39 +0000)]
test1 runs "cleanupEvent()" function by hand
Christopher Browne [Wed, 6 Feb 2008 20:51:56 +0000 (20:51 +0000)]
Suppress NOTIFY _%s_Event for subsequent SYNCs in a SYNC group
Christopher Browne [Wed, 6 Feb 2008 20:20:50 +0000 (20:20 +0000)]
Clean out code that used to generate "Confirm" events; the code was generally
commented out in the 1.2 branch; removing it now altogether.
Also, dropping out the "NOTIFY _%s_Confirm;" requests, as this may have been
causing extra unneccessary work.
Christopher Browne [Wed, 6 Feb 2008 18:04:02 +0000 (18:04 +0000)]
Increase SLON_DATA_FETCH_SIZE from 10 to 50, which was found to improve
performance on Large Update Data Sets.
Christopher Browne [Mon, 4 Feb 2008 20:56:22 +0000 (20:56 +0000)]
Comment on another reason (CONNECTION TIMEOUT) to separate a large
replication set into multiple pieces.
Christopher Browne [Tue, 29 Jan 2008 15:57:17 +0000 (15:57 +0000)]
updateReloid() does not need to raise an event
Christopher Browne [Tue, 29 Jan 2008 15:56:15 +0000 (15:56 +0000)]
Improve logging on mode switching between polling/listening
Christopher Browne [Tue, 29 Jan 2008 15:54:46 +0000 (15:54 +0000)]
Add #define to support 8.3 behaviour where typenameTypeId() takes 3 args
Christopher Browne [Mon, 28 Jan 2008 19:35:21 +0000 (19:35 +0000)]
Add a new discussion of how triggers function, as this changes pretty
massively in the next Slony-I release
Christopher Browne [Mon, 28 Jan 2008 19:34:25 +0000 (19:34 +0000)]
Add notes regarding requirement for CLONE NODE that there be
synchronization against SYNCs against other nodes
Christopher Browne [Mon, 28 Jan 2008 19:33:10 +0000 (19:33 +0000)]
SGML cleanup
Jan Wieck [Wed, 23 Jan 2008 17:12:34 +0000 (17:12 +0000)]
*** empty log message ***
Christopher Browne [Mon, 21 Jan 2008 19:27:37 +0000 (19:27 +0000)]
Add basic documentation, release notes on CLONE PREPARE/CLONE FINISH
Jan Wieck [Mon, 21 Jan 2008 18:54:11 +0000 (18:54 +0000)]
New even type CLONE_NODE.
This will allow to duplicate an existing subscriber using pg_dump or
even filesystem level of copying the database cluster. See the script
src/ducttape/test_1_clonenode for an example of how this works.
Documentation still to be done.
Jan
Christopher Browne [Wed, 9 Jan 2008 20:55:47 +0000 (20:55 +0000)]
Add in notes about C programs random_number, random_string
Christopher Browne [Wed, 9 Jan 2008 20:49:33 +0000 (20:49 +0000)]
Fix usability of C-based random data generators:
1. Change existence test to use -x (is program executable) rather than -e
(merely checking for existence)
2. Typo in string handling was throwing away string values when we ran
the C-based "random_string" program
Christopher Browne [Wed, 9 Jan 2008 20:47:48 +0000 (20:47 +0000)]
Minor spelling fix
Christopher Browne [Tue, 8 Jan 2008 20:43:24 +0000 (20:43 +0000)]
Change slonik build to query Postgres for the share directory at runtime
Per Dave Page
Christopher Browne [Fri, 4 Jan 2008 23:05:50 +0000 (23:05 +0000)]
Add notes to FAQ concerning findings for bug #27
Christopher Browne [Fri, 4 Jan 2008 19:52:35 +0000 (19:52 +0000)]
Per Bugzilla bug #30 observed by Peter Eisentraut <peter_e@gmx.net>
It says
Slony Binaries in: @@PGBINDIR@@
but it should say
Slony Binaries in: @@SLONBINDIR@@
because that is where they are actually installed.
Christopher Browne [Thu, 3 Jan 2008 15:48:49 +0000 (15:48 +0000)]
Note logging of configuration in release notes
Christopher Browne [Thu, 3 Jan 2008 15:47:21 +0000 (15:47 +0000)]
Added dump_configuration() function that dumps (to Slony-I log), at start
time, the values of all configuration parameters.
Example:
2008-01-03 15:44:55 UTC CONFIG main: Integer option vac_frequency = 3
2008-01-03 15:44:55 UTC CONFIG main: Integer option log_level = 0
2008-01-03 15:44:55 UTC CONFIG main: Integer option sync_interval = 2000
2008-01-03 15:44:55 UTC CONFIG main: Integer option sync_interval_timeout = 10000
2008-01-03 15:44:55 UTC CONFIG main: Integer option sync_group_maxsize = 20
2008-01-03 15:44:55 UTC CONFIG main: Integer option desired_sync_time = 60000
2008-01-03 15:44:55 UTC CONFIG main: Integer option syslog = 0
2008-01-03 15:44:55 UTC CONFIG main: Integer option quit_sync_provider = 0
2008-01-03 15:44:55 UTC CONFIG main: Integer option quit_sync_finalsync = 0
2008-01-03 15:44:55 UTC CONFIG main: Integer option sync_max_rowsize = 8192
2008-01-03 15:44:55 UTC CONFIG main: Integer option sync_max_largemem =
5242880
2008-01-03 15:44:55 UTC CONFIG main: Integer option remote_listen_timeout = 300
2008-01-03 15:44:55 UTC CONFIG main: Boolean option log_pid = 0
2008-01-03 15:44:55 UTC CONFIG main: Boolean option log_timestamp = 1
2008-01-03 15:44:55 UTC CONFIG main: Boolean option cleanup_deletelogs = 0
2008-01-03 15:44:55 UTC CONFIG main: Real option real_placeholder = 0.000000
2008-01-03 15:44:55 UTC CONFIG main: String option cluster_name = dw
2008-01-03 15:44:55 UTC CONFIG main: String option conn_info = dbname=dw
2008-01-03 15:44:55 UTC CONFIG main: String option pid_file = (null)
2008-01-03 15:44:55 UTC CONFIG main: String option log_timestamp_format = %Y-%m-%d %H:%M:%S %Z
2008-01-03 15:44:55 UTC CONFIG main: String option archive_dir = (null)
2008-01-03 15:44:55 UTC CONFIG main: String option sql_on_connection = (null)
2008-01-03 15:44:55 UTC CONFIG main: String option lag_interval = (null)
2008-01-03 15:44:55 UTC CONFIG main: String option command_on_logarchive = (null)
2008-01-03 15:44:55 UTC CONFIG main: String option syslog_facility = LOCAL0
2008-01-03 15:44:55 UTC CONFIG main: String option syslog_ident = slon
2008-01-03 15:44:55 UTC CONFIG main: String option cleanup_interval = 10 minutes
2008-01-03 15:44:55 UTC CONFIG slon: worker process created - pid = 28482
Christopher Browne [Wed, 2 Jan 2008 19:12:42 +0000 (19:12 +0000)]
Indicate changes to slon policy resulting from the recent patch that allows
configuring slon to never DELETE from sl_log_? tables, but to rather solely
trim data via TRUNCATE.
Notably this includes:
- Documenting the new slon parameters
- Changing wordings in several places to indicate the behaviour change
Christopher Browne [Wed, 2 Jan 2008 19:00:27 +0000 (19:00 +0000)]
Apply changes to cleanup thread:
- by default, use TRUNCATE rather than DELETE to trim sl_log_? tables
- moved substantially all logic for trimming of old data out of the
C code, and into cleanupThread() stored function
The stored function now can be used to request cleanup, so that we
might force it to happen at an unexpected time, such as manually,
for testing, as part of a test script.
- added parameters to slon to support this:
interval - cleanup_interval (default '10 minutes')
This controls how quickly old events are trimmed out. It used to
be a hard-coded value.
Old events are trimmed out once the confirmations are aged by
(cleanup_interval).
This then controls when the data in sl_log_1/sl_log_2 can be dropped.
Data in *those* tables is deleted when it is older than the
earliest XID still captured in sl_event.
boolean - cleanup_deletelogs (default 'false')
This controls whether or not we DELETE data from sl_log_1/sl_log_2
By default, we now NEVER delete data from the log tables; we
instead use TRUNCATE.
David Fetter [Thu, 20 Dec 2007 00:56:46 +0000 (00:56 +0000)]
Updated copyright notices to 2007.
Devrim GUNDUZ [Tue, 18 Dec 2007 20:54:11 +0000 (20:54 +0000)]
Use bzip2 for building tarballs for rpms builds
Christopher Browne [Mon, 17 Dec 2007 20:18:31 +0000 (20:18 +0000)]
Augment failover docs to make it clearer what is involved in refreshing
DB connections after failover/MOVE SET.
Christopher Browne [Thu, 13 Dec 2007 17:19:28 +0000 (17:19 +0000)]
Per bug #23
http://www.slony.info/bugzilla/show_bug.cgi?id=23
Need to ensure that the config parameter, sync_group_maxsize, is actually
used to constrain the maximum number of SYNCs grouped together.
Also changed debugging level from SLON_DEBUG3 to SLON_DEBUG2, for the
nearby log statement. (for 1.2 branch; SLON_DEBUG1, for HEAD)
Christopher Browne [Tue, 11 Dec 2007 21:27:31 +0000 (21:27 +0000)]
Document the problem observed in Bug #25
http://www.slony.info/bugzilla/show_bug.cgi?id=25
Christopher Browne [Tue, 11 Dec 2007 21:27:05 +0000 (21:27 +0000)]
Fix tagging errors
Christopher Browne [Tue, 11 Dec 2007 20:39:41 +0000 (20:39 +0000)]
No longer need make_function_strict() stored procedure as of 8.3
Christopher Browne [Tue, 11 Dec 2007 20:37:39 +0000 (20:37 +0000)]
Remove code that fiddles xxidin() function that is no longer needed as of
PG 8.3
Christopher Browne [Tue, 11 Dec 2007 20:32:50 +0000 (20:32 +0000)]
Change query looking for O/S sl_log_1 data to accomodate the change from
using "xxid" functions/types to the new snapshot function in PG 8.3.
Christopher Browne [Tue, 11 Dec 2007 20:30:58 +0000 (20:30 +0000)]
Change comment about ev_minxxid, which has been eliminated from sl_event
Christopher Browne [Tue, 11 Dec 2007 20:27:11 +0000 (20:27 +0000)]
Note changes resulting from new txid functions in PG 8.3
Jan Wieck [Tue, 11 Dec 2007 19:30:30 +0000 (19:30 +0000)]
Adjustments for PostgreSQL 8.3 and moving to the now builtin
txid data type.
Jan
Christopher Browne [Mon, 3 Dec 2007 23:53:42 +0000 (23:53 +0000)]
Update txid functions to indicate that they live in pg_catalog, not the
public schema.
Christopher Browne [Mon, 3 Dec 2007 16:29:10 +0000 (16:29 +0000)]
Add "rule/trigger search tool" to release notes
Christopher Browne [Mon, 3 Dec 2007 16:26:09 +0000 (16:26 +0000)]
Add in script to support usage described in bug #19
<http://www.slony.info/bugzilla/show_bug.cgi?id=19>, including
documentation in admin guide.
Christopher Browne [Thu, 29 Nov 2007 21:37:46 +0000 (21:37 +0000)]
Add notes to INIT CLUSTER about maximum sizing of cluster name
Christopher Browne [Thu, 29 Nov 2007 21:29:03 +0000 (21:29 +0000)]
Fix to bug #15
http://www.slony.info/bugzilla/show_bug.cgi?id=15
Fix to bug #15 - where long cluster name (>40 chars) leads to
things breaking when an index name is created that contains
the cluster name.
-> Warn upon creating a long cluster name.
-> Give a useful exception that explains the cause rather
than merely watching index creation fail.
Christopher Browne [Thu, 29 Nov 2007 15:57:56 +0000 (15:57 +0000)]
Change to libpq detection code...
- This version of Slony-I no longer needs to support versions earlier than
8.3, so we alter rules that related to earlier versions, either:
- Drop irrelevant tests
- Report that they are irrelevant
Christopher Browne [Tue, 27 Nov 2007 21:43:46 +0000 (21:43 +0000)]
Add in a fairly complex replication test involving multiple clusters that
coexist, 3 clusters across 4 distinct databases.
This provides some verification that some of the tools function properly.
Christopher Browne [Thu, 22 Nov 2007 23:02:06 +0000 (23:02 +0000)]
Remove TODO items now in Bugzilla items 11, 12, 13
Christopher Browne [Thu, 22 Nov 2007 22:58:42 +0000 (22:58 +0000)]
Remove items that are reflected in bugzilla bugs #9 and 10
Christopher Browne [Thu, 22 Nov 2007 22:56:21 +0000 (22:56 +0000)]
Remove TODO item tracked as bug #10 in Bugzilla
Christopher Browne [Thu, 22 Nov 2007 17:50:07 +0000 (17:50 +0000)]
Note change to tools/configure-replication.sh
Christopher Browne [Thu, 22 Nov 2007 17:47:03 +0000 (17:47 +0000)]
Add in SLONIKCONFDIR to allow the gentle user to specify a directory
in which to place the slonik files. This will enable using this tool
for automated processes.
Christopher Browne [Fri, 2 Nov 2007 15:31:43 +0000 (15:31 +0000)]
Per Steve Hindmarch - some missing newlines in configure-replication.sh
Christopher Browne [Fri, 26 Oct 2007 18:36:05 +0000 (18:36 +0000)]
Don't bother building slony_logshipper on Win32
- Per Hiroshi Saito
Christopher Browne [Fri, 26 Oct 2007 18:23:04 +0000 (18:23 +0000)]
From: Bernd Helmle <mailings@oopsware.de>
--On Donnerstag, Oktober 25, 2007 13:11:49 -0700 slony@estrider.com wrote:
> There is a flaw in the slonik_build_env.pl script. The print test
> should be >=1 and not > 1. It currently will not print databases where
> there are only single tables or sequences.
>
You're right. I've created a patch with your changes plus a fix to
repair the -schema option, which was broken, too, and cc'ed
slony1-patches so someone can review and apply it.
Note that slonik_build_environment is called slonik_build_env only,
the former was removed.
--
Thanks
Bernd
-------------------------------------------------
Looks good to me - applied to HEAD branch...
Christopher Browne [Fri, 26 Oct 2007 14:47:24 +0000 (14:47 +0000)]
Improvements to test for MERGE SET - better use of WAIT FOR EVENT,
notably.
Christopher Browne [Thu, 25 Oct 2007 18:32:00 +0000 (18:32 +0000)]
Oops - add in missing function that was added to 1.2 branch
Christopher Browne [Tue, 23 Oct 2007 17:00:27 +0000 (17:00 +0000)]
Add versions 1.2.11, 1.2.12 to HEAD's "update functions" set of versions
that it recognizes.
Christopher Browne [Tue, 23 Oct 2007 16:55:49 +0000 (16:55 +0000)]
- log shipping check - do a Restart Node if the request to update
the shared archive file number fails due to serialization
- reshaping of DDL handling code to get the changes recently made
to fix v1.2
Christopher Browne [Tue, 23 Oct 2007 16:52:56 +0000 (16:52 +0000)]
Impose "order by tab_id" in several places where tables are being
altered to turn replication on/off.
Seems preferable to order this way; that way we're actually using
the table ordering that the docs claim we use.
Christopher Browne [Tue, 23 Oct 2007 15:01:09 +0000 (15:01 +0000)]
Win32 test files per Hiroshi Saito
Christopher Browne [Mon, 22 Oct 2007 20:50:35 +0000 (20:50 +0000)]
Fix per David Rees
Christopher Browne [Mon, 22 Oct 2007 20:49:26 +0000 (20:49 +0000)]
Changes made in 1.2 branch - new log message, WAIT ON EVENT revisions
Christopher Browne [Mon, 22 Oct 2007 18:45:02 +0000 (18:45 +0000)]
Fixes to "merge set" test
Christopher Browne [Mon, 22 Oct 2007 18:06:38 +0000 (18:06 +0000)]
Add (to partitioning test) script for generating the "weak user" schema.
Christopher Browne [Fri, 19 Oct 2007 21:31:35 +0000 (21:31 +0000)]
Fix up subscription WAIT FOR EVENT requests so that they use SYNCs in
between and _properly_ wait for subscriptions to complete.
Jan Wieck [Fri, 19 Oct 2007 18:38:35 +0000 (18:38 +0000)]
Move slony to use the new txid contrib module instead of xxid.
Jan
Christopher Browne [Fri, 19 Oct 2007 15:22:35 +0000 (15:22 +0000)]
Lots of changes to log shipping test based on the changes introduced
in 1.2.12
Christopher Browne [Fri, 19 Oct 2007 15:21:46 +0000 (15:21 +0000)]
Changes to tests to support log shipping better
Christopher Browne [Fri, 19 Oct 2007 15:20:05 +0000 (15:20 +0000)]
Clean up formatting of README for test
Christopher Browne [Fri, 19 Oct 2007 15:19:34 +0000 (15:19 +0000)]
Inheritance test now does some subtransactions with rollback, with
some SLEEP statements to encourage potential for locking
Christopher Browne [Thu, 4 Oct 2007 15:33:45 +0000 (15:33 +0000)]
Add newlines to logging statements that were getting combined
with the previous line
Christopher Browne [Thu, 4 Oct 2007 15:32:31 +0000 (15:32 +0000)]
make "hostname" arguments more portable (after Mac OS-X testing)
Christopher Browne [Thu, 4 Oct 2007 15:31:44 +0000 (15:31 +0000)]
Add a negative test to DDL script test: try to run some wrong SQL
Christopher Browne [Thu, 4 Oct 2007 15:30:34 +0000 (15:30 +0000)]
Strip quotes from PID file name in slon.conf file
Christopher Browne [Thu, 4 Oct 2007 15:29:41 +0000 (15:29 +0000)]
Ran autoconf to generate fresh configure
Christopher Browne [Fri, 28 Sep 2007 20:24:35 +0000 (20:24 +0000)]
Change variable name 'node' to 'tnode' when processing DDL on individual
nodes... This seemed to be getting aliased by another 'node' somewhere.
Jan Wieck [Thu, 27 Sep 2007 18:02:52 +0000 (18:02 +0000)]
Let the logshipper wait on -t and -T until the running daemon has
destroyed the semaphore set.
Jan
Jan Wieck [Thu, 27 Sep 2007 14:22:59 +0000 (14:22 +0000)]
Fixing the archive sequence generations. All non-SYNC events must
start the local transaction before creating the archive as well, so
that the lock on the archive counter table serializes archive creation.
Test 8 is getting expanded bit by bit into a mega-test.
Jan
Christopher Browne [Wed, 26 Sep 2007 22:19:57 +0000 (22:19 +0000)]
Add in a test that combines EXECUTE SCRIPT with an attempt at
deadlocking the subscriber node.
To try to help resolve issues Jeff Frost ran into...
Christopher Browne [Tue, 25 Sep 2007 21:36:29 +0000 (21:36 +0000)]
Modify "testinherit" test to add in a test of subtransactions/savepoints.
- A bunch of concurrent psql sessions generate random lists of regions/products
- each of them performs purchase transactions based on these lists
- some of the purchases get rolled back
- some "sleep" requests go in in order to increase concurrency
Christopher Browne [Tue, 25 Sep 2007 18:25:11 +0000 (18:25 +0000)]
Add in a test that exercises MERGE SET; this is based on the partitioning
test (tests/testpartition), the difference being that in this test, the
partitions are added to replication via:
- creating a new set (#999)
- adding the new table to that set
- subscribing all nodes to that set
- merging set #999 into set #1
Christopher Browne [Mon, 24 Sep 2007 21:26:34 +0000 (21:26 +0000)]
Add in test that exercises the conditions noticed in the following discussion
thread:
http://lists.slony.info/pipermail/slony1-general/2007-September/006642.html
The problem takes place when a node receives two different sets which
have the same origin from two different providers. This ought to be
supported, but does not presently work.
Christopher Browne [Mon, 24 Sep 2007 21:07:45 +0000 (21:07 +0000)]
Add a script to help duplicate a node.
Christopher Browne [Fri, 21 Sep 2007 22:11:14 +0000 (22:11 +0000)]
Plans for more tests
Christopher Browne [Fri, 21 Sep 2007 22:10:54 +0000 (22:10 +0000)]
Document what tests have recently gotten created in NG tests
Christopher Browne [Fri, 21 Sep 2007 21:53:40 +0000 (21:53 +0000)]
Revise NG tests to exercise Slonik keywords DEFINE and INCLUDE:
- All tests now 'include<$tmp/slonik.preamble>;' to draw in cluster info
which includes a DEFINE for each node
- testlistenpath has been changed so that all Slonik references to nodes
use @node${number} to reference a DEFINEd value
This somewhat obsolesces Duct Tape Test #7
Christopher Browne [Fri, 21 Sep 2007 21:34:56 +0000 (21:34 +0000)]
Add in a test that has two disjoint sets to make sure that events
are propagating across the cluster
Christopher Browne [Mon, 17 Sep 2007 22:12:20 +0000 (22:12 +0000)]
Fix syslog option handling (via adding a #ifdef-wrapped parameterized macro)...
Several syslog options did not have any option remapping Slony-I debug levels
to syslog levels, with the result that those levels got mapped, by syslog, to
being ERRORs, which wasn't valid.