Luiz K. Matsumura [Fri, 26 Oct 2012 20:56:35 +0000 (18:56 -0200)]
Init script sample using altperl routines
Example of a start/stop service script using altperl and a
slon config file example
Luiz K. Matsumura [Fri, 26 Oct 2012 20:54:03 +0000 (18:54 -0200)]
slony_show_configuration can now return cluster name property
Luiz K. Matsumura [Fri, 26 Oct 2012 20:53:13 +0000 (18:53 -0200)]
New Watchdog options - verbosity
Implementing new parameters on slon_tools.conf
WATCHDOG_VERSION : select watchdog to start
1: slon_watchdog
2: slon_watchdog2
WATCHDOG_VERBOSE: enable/disable verbosity of watchdog logs
START_WATCHDOG and SLEEP_TIME: just declaring since the funcionality
was already implemented
slon_watchdog2: making the routine functional since some bugs was detected,
routine updating as changes detected on slon_watchdog
Luiz K. Matsumura [Fri, 26 Oct 2012 20:34:20 +0000 (18:34 -0200)]
Implementing capability of kill only a node on slon_kill
New parameter to kill only processes of one node
Refactoring to create subroutines.
Adjust of commands to grant that they will be sorted by pid values
Luiz K. Matsumura [Fri, 26 Oct 2012 20:19:52 +0000 (18:19 -0200)]
Change get_pid to return the smallest pid
get_pid changed to return the smallest pid of processes that match
the regex, probably the main proccess
Luiz K. Matsumura [Fri, 26 Oct 2012 11:19:18 +0000 (09:19 -0200)]
Identify node on log messages generated by slon_watchdog
Luiz K. Matsumura [Fri, 26 Oct 2012 11:18:33 +0000 (09:18 -0200)]
Adjusts slon_kill command
Fix slon_kill regex to match new command format in slon_start
Use slon_tools.conf to identify daemons of current cluster
Luiz K. Matsumura [Fri, 26 Oct 2012 11:14:46 +0000 (09:14 -0200)]
Fixing regex expression get_pid
Luiz K. Matsumura [Thu, 25 Oct 2012 19:34:17 +0000 (17:34 -0200)]
Allowing get_pid to accept more generic format of parameter nodeNumber
change regex to get_pid accept "node1" or "1" as parameter to identify node number
Luiz K. Matsumura [Thu, 25 Oct 2012 16:45:02 +0000 (14:45 -0200)]
Drop get_node_name from slon-tools.pm
The function get_node_name seen meaningless...
Changed slon_status to verify if the node is defined by using $DBNAME array
that is a obligatory value.
Luiz K. Matsumura [Thu, 25 Oct 2012 14:12:36 +0000 (12:12 -0200)]
Start slon passing with a slon configuration file
Implement a new parameter on node configuration to allow to specify
a slon configuration file to be used to start the slon daemon
adjusted slon_start and get_pid functions to consider the new start string
Luiz K. Matsumura [Thu, 25 Oct 2012 13:07:19 +0000 (11:07 -0200)]
implementing PGPASSWORD workaround in query_slon_status
If password is defined to node use PGPASSWORD environment variable to allow psql connect with the database
Luiz K. Matsumura [Thu, 25 Oct 2012 12:58:54 +0000 (10:58 -0200)]
revision of node_is_subscribing function
- Type cast bug
- Space on qq and { causing sintax error
- initializing $nodenum parameter
- Creating a workaround in case of password for conection is defined to allow psql to connect
Used PGPASSWORD enviroment variable.
Luiz K. Matsumura [Wed, 24 Oct 2012 20:53:09 +0000 (18:53 -0200)]
slonik_create_set : adding set name on comment of set
Added the unique name of set on the comment of create set comand
to help when inspecting the sets under postgres
Luiz K. Matsumura [Wed, 24 Oct 2012 20:34:59 +0000 (18:34 -0200)]
Changing behaviour of create a subdirectory slony1 under $LOGDIR
Don't create subdirectory slony1 under $LOGDIR to avoid create
a unnecessary depth since we can set $LOGDIR with an appropriate path
Luiz K. Matsumura [Wed, 24 Oct 2012 20:14:03 +0000 (18:14 -0200)]
slony_show_configuration.pl : Implementing possibility to return only a especific property on configuration of node
Steve Singer [Fri, 31 Aug 2012 13:14:31 +0000 (09:14 -0400)]
updating release to 2.2.0 b2
Steve Singer [Thu, 30 Aug 2012 18:49:33 +0000 (14:49 -0400)]
use numeric error code for detecting unique violations on sl_nodelock
The text of the error message has changed slightly in different postgresql
versions, ie 8.3+. Using the numeric error code will protect us from this.
This fixes bug 276.
Steve Singer [Thu, 23 Aug 2012 17:03:05 +0000 (13:03 -0400)]
fixed bug 275 - memory leak in monitor thread
Patch from Ulrich Weber <ulrich.weber@sophos.com>
Steve Singer [Mon, 13 Aug 2012 11:47:42 +0000 (07:47 -0400)]
bug 264 - second part.
This part of the patch was submitted by paulro@yandex.ru
Steve Singer [Fri, 10 Aug 2012 20:50:39 +0000 (16:50 -0400)]
updating release notes, bug 272
Steve Singer [Fri, 10 Aug 2012 20:48:02 +0000 (16:48 -0400)]
Merge branch 'bug272'
Steve Singer [Fri, 10 Aug 2012 20:46:40 +0000 (16:46 -0400)]
checking for insufficient memory
Steve Singer [Tue, 7 Aug 2012 16:58:30 +0000 (12:58 -0400)]
Bug 258.
Obtain an access exclusive lock not an exclusive one.
A concurrent select statement does not seem to block
the exclusive lock (but does block the access exclusive one)
and will block the truncate.
Steve Singer [Tue, 7 Aug 2012 16:54:20 +0000 (12:54 -0400)]
Updating release notes
Christopher Browne [Wed, 25 Apr 2012 20:58:40 +0000 (17:58 -0300)]
Possibly resolves bug slony.info/bugzilla/show_bug.cgi?id=264 where log_actionseq values are being generated in random-ish order, which means that subscriptions that take a long, long time tend to have a LOT of log_actionseq values to exclude, which builds into a query so large that it blows out the Postgres query parser.
Imposing by-log_actionseq order should rectify this.
Steve Singer [Wed, 25 Jul 2012 19:01:27 +0000 (15:01 -0400)]
Release notes update
Steve Singer [Sun, 22 Jul 2012 01:42:03 +0000 (21:42 -0400)]
Do not delete schemadoc.xml as part of a distclean.
We will include schemadoc.xml in the tar file as has been traditionally
done in slony 2.0.x. This allows the documentation to be built on machines
that don't have Postgresql running.
Steve Singer [Fri, 20 Jul 2012 15:55:59 +0000 (11:55 -0400)]
Adding in documentation section on the KEY option 'set add table'.
This was improperly removed from the documentation
Steve Singer [Thu, 5 Jul 2012 12:12:36 +0000 (08:12 -0400)]
Deal with embedded single quotes inside IN_STRING.
This is part of bug272
Steve Singer [Wed, 4 Jul 2012 15:46:04 +0000 (11:46 -0400)]
Modifying unit test to trigger bug 272
Steve Singer [Wed, 4 Jul 2012 12:59:53 +0000 (08:59 -0400)]
Fixing bug 272.
Pointers into yytext can't be stored and reused in a later rule
yy_get_next_buffer() might invalidate that pointer.
Instead we allocate our own buffer to store the IN_STRING state.
Steve Singer [Thu, 21 Jun 2012 15:47:22 +0000 (11:47 -0400)]
Fixes for Win32/MS Visual C.
Gets slony 2.2 compiling on VC and associated documentation updates.
Steve Singer [Wed, 20 Jun 2012 17:16:04 +0000 (13:16 -0400)]
Updating release notes
Christopher Browne [Wed, 20 Jun 2012 15:32:09 +0000 (11:32 -0400)]
Add notes on how to upgrade from 2.1 to 2.2
Christopher Browne [Wed, 20 Jun 2012 14:52:57 +0000 (10:52 -0400)]
Revise release note formatting, add reference to bug #137
Steve Singer [Wed, 20 Jun 2012 14:04:02 +0000 (10:04 -0400)]
Fixing debug message & typo
Steve Singer [Mon, 18 Jun 2012 18:52:57 +0000 (14:52 -0400)]
Updating documentation to reflect the new FAILOVER behaviour
Steve Singer [Mon, 18 Jun 2012 17:26:24 +0000 (13:26 -0400)]
Fixing documentation build errors
Steve Singer [Mon, 18 Jun 2012 13:38:41 +0000 (09:38 -0400)]
Updating version number to beta1
Steve Singer [Mon, 18 Jun 2012 13:20:28 +0000 (09:20 -0400)]
Bug 270.
Perform a TRUNCATE .. CASCADE on the replica when we replicate a truncate command.
This is how 2.1 works
Steve Singer [Tue, 12 Jun 2012 18:33:21 +0000 (14:33 -0400)]
Bug 263
Normally, one should be able to do
./configure --flags ... CPPFLAGS='...' CFLAGS='...' LDFLAGS='...'
With slony, this works for CFLAGS, but CPPFLAGS and LDFLAGS are not recorded in
Makefile.global. If this were added, some assignments in Makefile.global
because redundant, because the variable already contains the relevant values
from configure.
Patch from Peter Eisentraut
Steve Singer [Wed, 30 May 2012 14:07:00 +0000 (10:07 -0400)]
Fix compile errors/warnings when compiling against pg 9.2
Steve Singer [Wed, 30 May 2012 14:04:34 +0000 (10:04 -0400)]
Bug 268
Postgresql 9.2 (and later) has changed procpid to be pid in pg_stat_activity.
Use the newer column name when connected to a new pgsql
Steve Singer [Wed, 30 May 2012 21:47:53 +0000 (17:47 -0400)]
Additional remerges for bug 248
Steve Singer [Wed, 30 May 2012 20:47:18 +0000 (16:47 -0400)]
Remerging bug 248.
This was partially unmerged as part of the COPY protocol commits
Stop tracking the sequence name on log shipping targets.
Instead include the sequence name in the log shipping files on
sequence updates. This avoids the issue of us not adding
new tracking table entries on an enable subscription.
Add an OMIT COPY option to the script to allow
a log shipping target to be upgraded to a new slony version
without reloading all the data. During an outage you
would drop the $clustername schema from the target
then run slony1_dump.sh with -omit_copy to reinit
the node.
Jan Wieck [Wed, 30 May 2012 20:15:56 +0000 (16:15 -0400)]
Merge branch 'master' of ssh://git.postgresql.org/slony1-engine into bug265
Christopher Browne [Wed, 16 May 2012 17:01:21 +0000 (13:01 -0400)]
Add in FAQ entry listing various ways to start slon
Christopher Browne [Wed, 16 May 2012 15:53:34 +0000 (11:53 -0400)]
Jeff Frost contributed a sample Upstart script; added this to the
shared configuration directory, along with a section in the admin
scripts page on documentation.
Jan Wieck [Tue, 15 May 2012 15:16:36 +0000 (11:16 -0400)]
Fix for bug 265. PostgreSQL 8.3 does not support the
array[]::text[] syntax for creating an empty text array.
All PG versions from 8.3 to 9.1 however do support
the syntax '{}'::text[] for doing the same.
Steve Singer [Wed, 2 May 2012 15:30:09 +0000 (11:30 -0400)]
Fixing syntax error in the example.
We use 'cluster name' not 'clustername'.
Steve Singer [Fri, 30 Mar 2012 12:28:56 +0000 (08:28 -0400)]
Fixing documentation error (The clone finish page
used 'clone prepare' in the synopsis)
Christopher Browne [Wed, 7 Mar 2012 17:37:11 +0000 (12:37 -0500)]
Ran pgindent against the codebase:
pgindent /var/lib/postgresql/postgresql/src/tools/pgindent/typedefs.list src/*/*.{c,h}
Christopher Browne [Mon, 5 Mar 2012 21:39:56 +0000 (16:39 -0500)]
COPY protocol makes some slon.conf parameters obsolete, so they should be dropped from regression test suite
Jan Wieck [Wed, 29 Feb 2012 19:56:24 +0000 (14:56 -0500)]
Merge remote branch 'wieck/master'
Jan Wieck [Wed, 29 Feb 2012 19:23:36 +0000 (14:23 -0500)]
Merge remote branch 'origin/master'
Jan Wieck [Wed, 29 Feb 2012 19:21:55 +0000 (14:21 -0500)]
Fixed a double increment bug in the statement scanner for execute script.
Steve Singer [Wed, 29 Feb 2012 16:03:58 +0000 (11:03 -0500)]
Adding dstring_terminate() before dstring_data()
dstring_append doesn't add a terminating null so
we need to explicitly do that before getting at the
data.
Christopher Browne [Mon, 27 Feb 2012 20:39:03 +0000 (15:39 -0500)]
Merge branch 'master' of ssh://git.postgresql.org/slony1-engine
Christopher Browne [Mon, 27 Feb 2012 20:38:53 +0000 (15:38 -0500)]
Add COPY Protocol notes to release notes
Steve Singer [Tue, 24 Jan 2012 19:31:51 +0000 (14:31 -0500)]
Updating documentation for the RESUBSCRIBE NODE command.
Steve Singer [Mon, 23 Jan 2012 22:05:48 +0000 (17:05 -0500)]
Implementing the RESUBSCRIBE NODE command.
Jan Wieck [Thu, 23 Feb 2012 14:38:16 +0000 (09:38 -0500)]
Increase sync timeout in MoveSet.
Jan Wieck [Thu, 16 Feb 2012 15:16:00 +0000 (10:16 -0500)]
Reintroduce ddlScript_complete() at the end of EXECUTE SCRIPT
to fix relnames and log trigger arguments.
Minor changes according to code review by Steve.
Jan Wieck [Wed, 15 Feb 2012 15:22:20 +0000 (10:22 -0500)]
Merge remote branch 'origin/master' into copy-protocol-new
Conflicts:
src/backend/slony1_funcs.sql
src/slon/remote_worker.c
Steve Singer [Fri, 3 Feb 2012 14:50:29 +0000 (09:50 -0500)]
Fixing merge errors
Steve Singer [Thu, 2 Feb 2012 16:49:17 +0000 (11:49 -0500)]
Significant changes to failover.
We now have explicit support for having multiple nodes FAILOVER at the same time.
The FAILOVER command will take a list of these nodes.
The DROP NODE command can also take a list of multiple nodes that get dropped
at once.
The FAILOVER code changes with this commit.
You can now only failover to nodes that are a failover candidate
as listed in the sl_failover_targets view. This nodes must have paths
to all the nodes the failed node has paths to and be a direct
subscriber to all sets the failed node is an origin for. See
the patch + documentation changes for a complete list of restrictions.
These restrictions were put in place to avoid complications in
getting a complete path+listen network where a non-direct subscriber
might be the most ahead node.
The FAILOVER command now promotes the most ahead node to be a master
and uses MOVE SET to make the desired backup node the new master.
The FAILOVER_NODE event replaces the FAILOVER_SET event and handles all
sets on that node. The event now comes from the most-ahead node using
the most ahead node's event_id events from the old origin are no longer faked.
This works because of the restrictions discussed above.
Steve Singer [Thu, 19 Jan 2012 21:21:10 +0000 (16:21 -0500)]
Updating spec file to support the version number in the
share/lib directory files.
Jan Wieck [Mon, 23 Jan 2012 20:57:54 +0000 (15:57 -0500)]
Merge remote branch 'ssinger/copy-protocol-new' into copy-protocol-new
Jan Wieck [Mon, 23 Jan 2012 20:55:49 +0000 (15:55 -0500)]
Merge remote branch 'origin/master' into copy-protocol-new
Jan Wieck [Mon, 23 Jan 2012 16:00:14 +0000 (11:00 -0500)]
Fixes for bug #260 by ssinger.
Steve Singer [Mon, 23 Jan 2012 15:52:44 +0000 (10:52 -0500)]
Merge remote branch 'jwieck/copy-protocol-new' into copy-protocol-new
Steve Singer [Mon, 23 Jan 2012 15:50:54 +0000 (10:50 -0500)]
Updating tests to work with the new EXECUTE SCRIPT syntax
Updating tests to no longer expect slony to allow multiple subscription
routes from the same origin to a particular receiver.
Steve Singer [Mon, 23 Jan 2012 13:30:22 +0000 (08:30 -0500)]
Updating EXECUTE SCRIPT syntax, removing set id.
Jan Wieck [Sat, 21 Jan 2012 16:59:01 +0000 (11:59 -0500)]
Adding log apply statistics.
Jan Wieck [Fri, 20 Jan 2012 23:00:42 +0000 (18:00 -0500)]
No longer need md5.h
Jan Wieck [Fri, 20 Jan 2012 21:17:06 +0000 (16:17 -0500)]
Change the apply trigger query cache to use a small footprint of the query
as the hash key. This key can be computed without doing any of the column
type lookups, making the query cache more efficient. If a prepared plan
is found, none of the catalog lookups are required at all.
Add a configuration option "apply_cache_size" to slon.conf.
Use custom wrappers around hash_any() and strcpm() to eliminate the use of md5.
Steve Singer [Fri, 20 Jan 2012 18:04:06 +0000 (13:04 -0500)]
Updating the tests to work with the changed EXECUTE SCRIPT behaviour.
'set id' is no longer passed to execute script.
Added a test to send multiple EVENT NODE nodes at once.
Steve Singer [Fri, 20 Jan 2012 14:52:08 +0000 (09:52 -0500)]
remove SET ID as an option from EXECUTE SCRIPT
since we no longer pay attention to it.
Jan Wieck [Thu, 19 Jan 2012 20:49:00 +0000 (15:49 -0500)]
Change format of the sl_log_1/2 tables so that they contain the
table name and schema as separate fields. Change the column names and
values from a rudimentary SQL statement into a text array.
Change the transport to use COPY protocol and move the part that does
the updates to user tables into a trigger that fires on sl_log_1/2.
Move DDL from sl_event into a separate sl_log_script table that is
pulled into the log selection so that schema changes flow at the
correct time in between data updates, not as separate events.
Steve Singer [Tue, 10 Jan 2012 16:48:53 +0000 (11:48 -0500)]
Bug #259.
Quote the sequence name when inserting a truncate event
into sl_log. This will make truncate triggers work for
installs with a mixed/upper case Slony clustername.
(cherry picked from commit
2c4f4061bee5f18ed7ce0d08af21dc850fe23a57)
Steve Singer [Mon, 9 Jan 2012 21:32:05 +0000 (16:32 -0500)]
MSVC specific changes for bug #256
Steve Singer [Thu, 5 Jan 2012 17:56:49 +0000 (12:56 -0500)]
Fixing bug introduced in SYNC group changes
commit
e0301ce99d8385b592ae5753572f6527fa4c447b
Steve Singer [Mon, 9 Jan 2012 14:51:43 +0000 (09:51 -0500)]
Additional fixes for bug #256.
-Fixing typo in variable name.
-Adding the defines to config.h.in
Steve Singer [Mon, 9 Jan 2012 14:33:10 +0000 (09:33 -0500)]
The slony1_funcs.so shared library and the .sql files slony installs
are now versioned when they are installed. This allows
multiple versions of Slony to be installed at the same time
in the same PostgreSQL lib/share directory.
Christopher Browne [Tue, 3 Jan 2012 21:21:50 +0000 (16:21 -0500)]
Bug #256
Check parms of set_config_option(), which have changed in 9.2
Christopher Browne [Fri, 23 Dec 2011 18:01:02 +0000 (13:01 -0500)]
Merge branch 'master' of ssh://git.postgresql.org/slony1-engine
Christopher Browne [Fri, 23 Dec 2011 18:00:48 +0000 (13:00 -0500)]
Fix SGML tagging
Steve Singer [Wed, 21 Dec 2011 22:11:22 +0000 (17:11 -0500)]
Add correct ps invocation for Darwin-based OSes
Steve Singer [Wed, 14 Dec 2011 15:34:35 +0000 (10:34 -0500)]
Fixing typo/missing data in the sample conf file
Christopher Browne [Tue, 13 Dec 2011 22:10:20 +0000 (17:10 -0500)]
Merge branch 'master' of git://git.postgresql.org/git/slony1-engine
Steve Singer [Tue, 13 Dec 2011 16:49:14 +0000 (11:49 -0500)]
Bug 255.
Changes for 9.1 to avoid pivot conflicts.
specify SERIALIZABLE for read only deferred transactions.
Make remote_worker calls to the provider read-only deferred
Make remote_worker calls on the local con READ COMMITTED.
Make the remote listener run in a READ ONLY DEFERRED transaction on
Based on a patch by Simon Riggs.
Christopher Browne [Mon, 12 Dec 2011 22:42:31 +0000 (17:42 -0500)]
Add release note on #240
Christopher Browne [Mon, 12 Dec 2011 22:40:53 +0000 (17:40 -0500)]
Bug #240 - improve vacuum documentation
Christopher Browne [Tue, 29 Nov 2011 17:02:59 +0000 (12:02 -0500)]
Merge branch 'master' of git://git.postgresql.org/git/slony1-engine
Steve Singer [Fri, 25 Nov 2011 18:24:19 +0000 (13:24 -0500)]
Updating URL to YUM repository to reflect yum.postgresql.org
Bug 251.
Christopher Browne [Fri, 25 Nov 2011 16:55:01 +0000 (11:55 -0500)]
Merge branch 'master' of git://git.postgresql.org/git/slony1-engine
Christopher Browne [Fri, 25 Nov 2011 16:42:17 +0000 (11:42 -0500)]
Bug #235 :: Redo SYNC GROUP SIZE logic
- Start at 1
- Double upon success, maxxing out based on the value of
sync_group_maxsize
- As we restart the remote worker thread upon failure, failure
resets group size back to 1
This gets rid of complex logic that had the false conceit that we had
any idea as to how long one might expect a SYNC to run, and could
reasonably extrapolate that to estimate long SYNC processing might take.
Steve Singer [Fri, 25 Nov 2011 14:18:23 +0000 (09:18 -0500)]
Fix for bug244.
Require a set id on create set.
Otherwise it will create set -1 which casues problems later on.
Steve Singer [Wed, 9 Nov 2011 14:13:37 +0000 (09:13 -0500)]
Fix for bug244.
Require a set id on create set.
Otherwise it will create set -1 which casues problems later on.