Toshihiro Kitagawa [Fri, 12 Aug 2011 16:37:22 +0000 (16:37 +0000)]
Fix the location which gets the parse tree because it is used after
the block. This fix is the following continuances:
[Pgpool-committers] pgpool - pgpool-II: It appears that the last commit
below has problem
Toshihiro Kitagawa [Fri, 12 Aug 2011 16:27:49 +0000 (16:27 +0000)]
Fix pcp_attach_node command so that it emits error message while
doing failover.
Toshihiro Kitagawa [Fri, 12 Aug 2011 06:46:04 +0000 (06:46 +0000)]
Fix bug which cannot use the cursors of JDBC driver on standby node.
In master/slave mode, the transaction commands come to be sent to
all nodes by this fix because the cursors of JDBC driver cannot use
without transactions.
But if the transaction command is "BEGIN READ WRITE" or "START TRANSACTION
READ WRITE", pgpool-II sends "BEGIN" instead of it to standby node.
The reason is because standby nodes disallow "READ WRITE" option.
Toshihiro Kitagawa [Fri, 12 Aug 2011 04:23:21 +0000 (04:23 +0000)]
Remove unnecessary function arguments.
Tatsuo Ishii [Thu, 11 Aug 2011 01:48:01 +0000 (01:48 +0000)]
Enhance release notes.
Tatsuo Ishii [Thu, 11 Aug 2011 01:47:41 +0000 (01:47 +0000)]
Enhance NEWS.
Tatsuo Ishii [Thu, 11 Aug 2011 00:15:12 +0000 (00:15 +0000)]
Fix description about backend_weight.
It can be changed by reloading pgpool.conf.
Toshihiro Kitagawa [Mon, 8 Aug 2011 07:08:10 +0000 (07:08 +0000)]
Fix typo.
Toshihiro Kitagawa [Mon, 8 Aug 2011 04:24:22 +0000 (04:24 +0000)]
Fix insert_lock so that it works correctly even if the table definition
such as the following:
CREATE TABLE login_log (id integer
DEFAULT nextval(('"login_log_id_seq"'::text)::regclass) NOT NULL
Tatsuo Ishii [Sun, 7 Aug 2011 11:05:39 +0000 (11:05 +0000)]
Remove unnecessary log.
Tatsuo Ishii [Sun, 7 Aug 2011 09:52:47 +0000 (09:52 +0000)]
Add sr_check_period, sr_check_user and sr_check_password.
These are used for streaming replication delay checking
and determining primary server.
Also fix get_config() which forget about health_check_password.
Tatsuo Ishii [Fri, 5 Aug 2011 10:53:55 +0000 (10:53 +0000)]
Update and sync those documents.
Tatsuo Ishii [Thu, 28 Jul 2011 07:19:40 +0000 (07:19 +0000)]
Allow to compile against PostgreSQL 9.1.
Toshihiro Kitagawa [Wed, 27 Jul 2011 04:03:43 +0000 (04:03 +0000)]
Add debug info.
Fix typo in error messages.
Toshihiro Kitagawa [Wed, 27 Jul 2011 03:32:13 +0000 (03:32 +0000)]
Fix bug with the handling of empty query.
JBoss sends empty queries for checking connections. We decided to replace
the empty query with dummy SELECT command not to affect load balance.
Empty query has been replaced with dummy DELETE command till now.
Note: The dummy command is not sent to backends.
See the following thread for more details:
[Pgpool-general] Confused about JDBC and load balancing
Tatsuo Ishii [Fri, 22 Jul 2011 12:20:47 +0000 (12:20 +0000)]
Add Simplified Chinese version of tutorial document
(doc/tutorial-zh_cn.html)
Contributed by Huang Bambo.
Tatsuo Ishii [Fri, 22 Jul 2011 11:06:41 +0000 (11:06 +0000)]
Update copyright year.
Tatsuo Ishii [Fri, 22 Jul 2011 11:01:36 +0000 (11:01 +0000)]
Simplified Chinese version of tutorial document.
Contributed by Huang Bambo.
Toshihiro Kitagawa [Fri, 22 Jul 2011 08:53:11 +0000 (08:53 +0000)]
Fix "Parse: Unable to parse the query:" log message so that it emits
the query. It has emitted the prepared statement name till now.
Tatsuo Ishii [Fri, 22 Jul 2011 00:22:33 +0000 (00:22 +0000)]
More fix suggested by Huang Bambo.
Tatsuo Ishii [Thu, 21 Jul 2011 23:19:32 +0000 (23:19 +0000)]
It appears that the last commit below has problem with the case
PREPARE(non extended protocol case) raises error, pointed out by
Toshihiro Kitagawa. ReadyForQuery() is changed to call
pool_unset_query_in_progress() whenver query_in_progress is set
(before it was only called when command_success is set).
Subject: [Pgpool-committers] pgpool - pgpool-II: Fix ProcessBackendResponse() so that it calls
From: t-ishii@pgfoundry.org (User T-ishii)
To: pgpool-committers@pgfoundry.org
Date: Wed, 20 Jul 2011 10:41:10 +0000 (UTC)
Log Message:
-----------
Fix ProcessBackendResponse() so that it calls pool_unset_command_success()
rather than pool_set_command_success() when received Error Response.
Modified Files:
--------------
pgpool-II:
pool_proto_modules.c (r1.99 -> r1.100)
(http://cvs.pgfoundry.org/cgi-bin/cvsweb.cgi/pgpool/pgpool-II/pool_proto_modules.c?r1=1.99&r2=1.100)
Tatsuo Ishii [Thu, 21 Jul 2011 23:01:41 +0000 (23:01 +0000)]
More fix suggested by Huang Bambo.
Tatsuo Ishii [Thu, 21 Jul 2011 09:16:54 +0000 (09:16 +0000)]
Enhance wording.
Tatsuo Ishii [Wed, 20 Jul 2011 10:41:10 +0000 (10:41 +0000)]
Fix ProcessBackendResponse() so that it calls pool_unset_command_success()
rather than pool_set_command_success() when received Error Response.
Tatsuo Ishii [Mon, 18 Jul 2011 07:33:43 +0000 (07:33 +0000)]
Make is_sequence_query, is_start_transaction_query,
is_strict_query to bool.
These functions have been already treated as bool
from callers and leaving these as int is sloppy.
So I changed these function prototypes.
Rename is_commit_query to is_commit_or_rollback_query. This is more
appropreate. Add is_commit_query(returns true if commit query) and
is_rollback_query(returns true if rollback query).
Toshihiro Kitagawa [Wed, 13 Jul 2011 06:10:33 +0000 (06:10 +0000)]
Prepare 3.1-alpha3.
Toshihiro Kitagawa [Wed, 13 Jul 2011 04:24:55 +0000 (04:24 +0000)]
Prepare 3.1-alpha3.
Toshihiro Kitagawa [Fri, 8 Jul 2011 05:54:01 +0000 (05:54 +0000)]
Prepare 3.1-alpha3.
Toshihiro Kitagawa [Thu, 7 Jul 2011 02:15:37 +0000 (02:15 +0000)]
Restructure samples of pgpool.conf.
- separate "ONLINE RECOVERY" section from "REPLICATION MODE" section
Because online recovery is used with streaming replication too.
- separate "HEALTH CHECK" and "FAILOVER AND FAILBACK" section from
"MASTER/SLAVE MODE" section
Because they are mode-independent.
- add backend_flagN directive.
- add "OTHERS" section for relcache_expire directive.
- rename "CONNECTION POOLING MODE" section to "CONNECTION POOLING"
- rename "PARALLEL AND QUERY CACHE MODE" section to
"PARALLEL MODE AND QUERY CACHE"
- replace "false" with "off" in comment.
Tatsuo Ishii [Wed, 6 Jul 2011 14:00:22 +0000 (14:00 +0000)]
Add health_check_password directive.
Enhance check_replication_time_lag() error message.
Patch contributed by Nicolas Thauvin.
His message:
We are currently not able to specify the password of the health check
user, which is needed at least for streaming replication lag. So I
added the possibility to specify it in the configuration file, in the
attached patch.
Also, I find the error message in check_replication_time_lag()
(pool_worker_child.c) not friendly :
pool_error("check_replication_time_lag: DB node is valid but no
persistent connection");
What about sending this error message to the debug level and adding a
second message on the error level? Fro example:
pool_error("check_replication_time_lag: could not connect to DB node %d,
check health check configuration", i);
Tatsuo Ishii [Tue, 5 Jul 2011 23:26:02 +0000 (23:26 +0000)]
Add uninstall SQL files to sql/. Patch contributed by
Nicolas Thauvin. Comments from him:
Please find attached uninstall_* SQL files for pgpool-recovery,
pgpool-regclass and pgpool-walrecrunning with the patch for the
Makefiles (against HEAD) needed to add them to PostgreSQL's contrib
directory.
Tatsuo Ishii [Tue, 5 Jul 2011 09:11:00 +0000 (09:11 +0000)]
Allow to show all process info using pcp_process_info, show pool_pools.
* it refactor "pcp processes info" related code to return all processes
* info from backend, similar to the "show pool_pools" query
* add pgpool process PID and backend id fields
* add the ability to show all these informations from pcp_proc_info
* command
* no change to the pcp_proc_info command default behavior
* new fields are added as last fields in the proc_info command only with
* the -a switch (both verbose or not)
Patch contributed by Jehan-Guillaume (ioguix) de Rorthais
Toshihiro Kitagawa [Tue, 5 Jul 2011 05:56:23 +0000 (05:56 +0000)]
Update configure.
It was created in conventional development environment(Vine Linux).
Tatsuo Ishii [Mon, 4 Jul 2011 01:32:59 +0000 (01:32 +0000)]
Fix uninitialized variable in failover() brought by previous commit.
Toshihiro Kitagawa [Fri, 1 Jul 2011 10:47:14 +0000 (10:47 +0000)]
Fix insert_lock so that it locks pgpool_catalog.insert_lock because
new PostgreSQL disallows a row lock against the sequence relation.
Add options for compatibility in configure script:
--enable-sequence-lock insert_lock compatible with pgpool-II 3.0 series
(until 3.0.4)
--enable-table-lock insert_lock compatible with pgpool-II 2.2 and 2.3
series
See following thread for more details:
[Pgpool-hackers] Alternative to locking sequences
Tatsuo Ishii [Wed, 29 Jun 2011 02:44:13 +0000 (02:44 +0000)]
Add new per backend directive "flag".
This controls per backend behavior. Currently "ALLOW_TO_FAILOVER" or
"DISALLOW_TO_FAILOVER" are allowed.
Tatsuo Ishii [Tue, 28 Jun 2011 23:07:41 +0000 (23:07 +0000)]
Allow time stamp rewriting to work with arbitrary expression in default
value of a column. Before we detected anything including now() then
simply replaced it to now(). This will lead to wrong rewriting of
default value. for example, timezone('utc'::text, now()).
Note that, however, this only adopts to simple queries. Extended protocols
(for example Java, PHP PDO) or sql "PREPARE" still remain same.
Tatsuo Ishii [Tue, 28 Jun 2011 22:48:26 +0000 (22:48 +0000)]
Allow time stamp rewriting to work with arbitrary expression in default
value of a column. Before we detected anything including now() then
simply replaced it to now(). This will lead to wrong rewriting of
default value. for example, timezone('utc'::text, now()).
Note that, however, this only adopts to simple queries. Extended protocols
(for example Java, PHP PDO) or sql "PREPARE" still remain same.
Guillaume Lelarge [Tue, 28 Jun 2011 12:31:47 +0000 (12:31 +0000)]
pgpool in raw mode is mistaken on alive nodes
VALID_BACKEND macro in raw mode should not be used. So, we add a specific
check when pgpool is running in raw mode.
Haruka harukat [Mon, 27 Jun 2011 03:05:47 +0000 (03:05 +0000)]
* add <a name> tags in the release note section
Tatsuo Ishii [Thu, 23 Jun 2011 06:38:33 +0000 (06:38 +0000)]
Add pgpool_adm functions. Contributed by Jehan-Guillaume (ioguix) de Rorthais.
Tatsuo Ishii [Wed, 22 Jun 2011 08:40:49 +0000 (08:40 +0000)]
Fix wrong log message of degenerate_backend_set(). Someone forgot to change
the message when doing copy and paste.
Toshihiro Kitagawa [Tue, 21 Jun 2011 07:24:00 +0000 (07:24 +0000)]
Fix bug which does not rewrite timestamp.
Fix rewrite_timestamp_walker so that it can process 2 cell list.
This used to work but was broken in 3.1.0 alpha2 using 9.0 parser.
Change history of functions to create lists in gram.y:
SystemFuncName() SystemTypeName()
8.1.3 original list_make2() list_make2()
8.1 import list_make1() list_make1()
8.3 import list_make1() list_make1()
8.4 import list_make1() list_make2()
9.0 import list_make2() list_make2()
Tatsuo Ishii [Sun, 19 Jun 2011 08:56:38 +0000 (08:56 +0000)]
Fix comments.
Toshihiro Kitagawa [Wed, 15 Jun 2011 01:31:27 +0000 (01:31 +0000)]
Fix typo.
Tatsuo Ishii [Sun, 12 Jun 2011 09:20:33 +0000 (09:20 +0000)]
Add --username(or -u) option to pg_md5. This allows to manage users
which do not have UNIX accounts. Patch contributed by Nicolas Thauvin.
Japanese doc change by Tatsuo Ishii.
Toshihiro Kitagawa [Wed, 8 Jun 2011 11:14:26 +0000 (11:14 +0000)]
Change "do_md5: read_password_packet failed" message into debug level.
This error message is emitted if frontends(psql etc.) authenticate
a password, because they might reconnect at the time.
Reason for change:
- PostgreSQL does not emit error messages at the time.
- pgpool-II does not emit error messages when using the other
authentication methods.
Toshihiro Kitagawa [Mon, 6 Jun 2011 08:18:55 +0000 (08:18 +0000)]
Fix typo.
Tatsuo Ishii [Sun, 5 Jun 2011 23:03:06 +0000 (23:03 +0000)]
Refactor ReadyForQuery(). Create close_standby_transactions() mainly
for cosmetic reason.
Toshihiro Kitagawa [Thu, 26 May 2011 05:37:25 +0000 (05:37 +0000)]
Fix log_per_node_statement so that it prints statements in the extended
query protocol. This used to work but was broken in 3.0.
Guillaume Lelarge [Wed, 25 May 2011 17:14:24 +0000 (17:14 +0000)]
Fix typo
Guillaume Lelarge [Mon, 23 May 2011 19:44:00 +0000 (19:44 +0000)]
Don't check replication lag if health check is disabled.
Tatsuo Ishii [Mon, 23 May 2011 10:55:36 +0000 (10:55 +0000)]
Fix bug with initializing pool_password in daemon mode.
Patch from Nicolas Thauvin.
To: pgpool-hackers@pgfoundry.org
Date: Tue, 17 May 2011 17:08:56 +0200
------------------------------------------------------------------------
When in daemon mode, connection hangs when using md5 authentication
method. The child process in charge of the client goes into an infinite
loop in pool_get_passwd() while trying to read the pool_passwd file.
The cause of the problem comes from the daemonize() function that closes
all file descriptors starting from 3, including the file descriptor of
the pool_passwd file. When pgpool gets to pool_get_passwd, the null
checks on the passwd_fd variable fail because the variable was not reset
to a NULL value.
Guillaume Lelarge [Mon, 23 May 2011 09:35:01 +0000 (09:35 +0000)]
backend status variable name was wrong. Replace the space with a _
Tatsuo Ishii [Fri, 13 May 2011 06:53:05 +0000 (06:53 +0000)]
Fix incorrect error message in pool_process_query().
It emits an error message that it timeouted in online recovery
even it is not in online recovery.
Tatsuo Ishii [Wed, 11 May 2011 01:51:36 +0000 (01:51 +0000)]
Add currval() and lastval() to black_function_list. If they are load
balanced, currval() or lastval() may be called before the result of
nextval() or setval() is propagated to slaves.
Tatsuo Ishii [Mon, 9 May 2011 23:37:14 +0000 (23:37 +0000)]
Fix cancel_request() so that it adopts the change made for 3.1
regarding UNIX domain socket name. Before we thought backend_port = ""
meant we should use UNIX domain socket, but now backend_port = "/"
means UNIX domain socket.
Tatsuo Ishii [Fri, 6 May 2011 23:43:26 +0000 (23:43 +0000)]
Oops. I removed too much.
Tatsuo Ishii [Fri, 6 May 2011 22:33:32 +0000 (22:33 +0000)]
Remove unnecessary logging at the end of parsing.
Tatsuo Ishii [Thu, 5 May 2011 02:39:14 +0000 (02:39 +0000)]
Fix incorrect calculation of LSN in bytes.
Change pg_last_xlog_receive_location() to
pg_last_xlog_replay_location() because the latter respects actually
replayed location which will give less chance for clients to have
stale data in standby.
Fix suggested by Anton Yuzhaninov.
Toshihiro Kitagawa [Tue, 3 May 2011 12:19:55 +0000 (12:19 +0000)]
Prepare 3.1-alpha2.
Toshihiro Kitagawa [Tue, 3 May 2011 09:18:01 +0000 (09:18 +0000)]
Prepare 3.1-alpha2.
Tatsuo Ishii [Mon, 2 May 2011 13:31:25 +0000 (13:31 +0000)]
Enhance online recovery in streaming replication mode.
Now restarting pgpool-II children is avoided when recovery finished.
So existing sessions can be continued while doing online recovery.
For this VALID_BACKEND macro is modified. pgpool-II main looks into
status in shared memory are as it were. pgpool-II child now looks
into local cache space. Pgpool-II child recognizes same number of
backends before failback util existing session ends. Then the
process exits with status 1, so that new process starts, which will
recognizes failback node. For this purpose "need_to_restart" flag
added to ProcessInfo structure.
Toshihiro Kitagawa [Mon, 2 May 2011 12:45:03 +0000 (12:45 +0000)]
Fix failover() so that it does not cause "kind does not match" error.
Add ifdef not to check pgpool_walrecrunning().
Adjust error messages.
Tatsuo Ishii [Sun, 1 May 2011 09:41:48 +0000 (09:41 +0000)]
Enhance online recovery documents in streaming replication.
Haruka harukat [Tue, 26 Apr 2011 11:33:19 +0000 (11:33 +0000)]
- add <a name="..."> tags into all parameter in the pgpool-ii user manual.
Toshihiro Kitagawa [Tue, 26 Apr 2011 00:44:36 +0000 (00:44 +0000)]
Add pool_process_reporting.h to be installed.
Tatsuo Ishii [Mon, 25 Apr 2011 22:58:04 +0000 (22:58 +0000)]
Fix bug with start_recoery().
It calls connect_backend() and then calls PQfinish(NULL) if
connect_backend() returns NULL, which is apparently wrong usage of
PQfinish(). Actually PQfinish() has been called in connect_backend()
in this case.
Also fix typo and enhance error message.
Tatsuo Ishii [Mon, 25 Apr 2011 13:56:46 +0000 (13:56 +0000)]
Add libpcp_ext.h to be installed.
Toshihiro Kitagawa [Mon, 25 Apr 2011 11:43:02 +0000 (11:43 +0000)]
Prepare 3.1-alpha2.
Toshihiro Kitagawa [Mon, 25 Apr 2011 11:33:37 +0000 (11:33 +0000)]
Forgot to add pool_process_reporting.h, pcp/pcp_pool_status.c.
Toshihiro Kitagawa [Mon, 25 Apr 2011 06:29:36 +0000 (06:29 +0000)]
Prepare 3.1-alpha2.
Tatsuo Ishii [Sat, 23 Apr 2011 07:49:02 +0000 (07:49 +0000)]
Refactor pcp related header files. Add libpcp_ext.h which is inteded to be
included by libpcp applications including PostgreSQL C functions.
Tatsuo Ishii [Sat, 23 Apr 2011 05:10:46 +0000 (05:10 +0000)]
Fix bug with pcp_connect() which causes double free.
Fix contributed by Jehan-Guillaume (ioguix) de Rorthais.
Toshihiro Kitagawa [Fri, 22 Apr 2011 11:40:37 +0000 (11:40 +0000)]
Increase relcache size from 32 to 128. This has the effect of reducing
"unnamed prepared statement does not exist" error.
Bug report from Joseph Jacob. See:
Subject: Fwd: Test java program
Date: Fri, 15 Apr 2011 12:16:04 -0700
Tatsuo Ishii [Fri, 22 Apr 2011 07:33:33 +0000 (07:33 +0000)]
Add pcp_pool_status, pcp_node_info.
Patch contributed by Jehan-Guillaume.
Toshihiro Kitagawa [Fri, 22 Apr 2011 02:50:27 +0000 (02:50 +0000)]
Prepare 3.1-alpha2.
Toshihiro Kitagawa [Fri, 22 Apr 2011 02:35:07 +0000 (02:35 +0000)]
Fix find_primary_node() so that it judges primary node correctly.
Remove pgpool_walrecrunning() from find_primary_node() and
add find_primary_node_repeatedly() that waits until finding primary node.
The reason for removing pgpool_walrecrunning() is that
"pg_is_in_recovery() AND pgpool_walrecrunning()" cannot judge
primary node and standby node correctly.
Toshihiro Kitagawa [Fri, 22 Apr 2011 01:09:03 +0000 (01:09 +0000)]
Add description about follow_master_command
Toshihiro Kitagawa [Thu, 21 Apr 2011 10:12:25 +0000 (10:12 +0000)]
Change follow_master_command so that it doesn't detach standby nodes,
if it's empty.
Toshihiro Kitagawa [Thu, 21 Apr 2011 05:00:52 +0000 (05:00 +0000)]
Remove unnecessary semicolon.
Toshihiro Kitagawa [Thu, 21 Apr 2011 02:44:01 +0000 (02:44 +0000)]
Add description of pcp_promote_node command.
Toshihiro Kitagawa [Thu, 21 Apr 2011 01:02:51 +0000 (01:02 +0000)]
Add pcp_promote_node command.
Patch contributed by Gilles Darold.
Tatsuo Ishii [Mon, 11 Apr 2011 22:55:44 +0000 (22:55 +0000)]
Add description about parameters for recovery_1st_stage_command and
recovery_2nd_stage_command. It's shame that the description
had been missing from day 0.
Tatsuo Ishii [Sat, 2 Apr 2011 23:41:51 +0000 (23:41 +0000)]
Remove unsed global variable "weight_master.
Toshihiro Kitagawa [Fri, 1 Apr 2011 10:57:35 +0000 (10:57 +0000)]
Fix hangup with CloseComplete(C) message handling.
Toshihiro Kitagawa [Thu, 31 Mar 2011 08:12:43 +0000 (08:12 +0000)]
Refactor pool_process_query().
- Fix hangup when received Flush(H) message.
See following for more details:
Subject: [Pgpool-general] Issue with pgpool-II-3.0.3-beta1 prepared
statements execution
Date: Tue, 22 Feb 2011 13:55:25 +0300
- Fix possible hangup that happen for the receiving timing of
ReadyForQuery(Z) message.
How to reproduce:
frontend -> pgpool-II -> pgpool-II -> backend
Both pgpool-II are in connection pooling mode.
Toshihiro Kitagawa [Wed, 30 Mar 2011 02:13:07 +0000 (02:13 +0000)]
Add follow_master_command directive.
We need to fix find_primary_node() that is still pending.
Patch contributed by Gilles Darold.
Toshihiro Kitagawa [Thu, 17 Mar 2011 02:12:09 +0000 (02:12 +0000)]
Fix warning of parser_set_param().
Toshihiro Kitagawa [Wed, 16 Mar 2011 09:11:25 +0000 (09:11 +0000)]
Forgot to add files.
Toshihiro Kitagawa [Wed, 16 Mar 2011 05:55:12 +0000 (05:55 +0000)]
Fix hangup when using /*NO LOAD BALANCE*/ comment.
The condition is load_balance_mode=true and PRIMARY_NODE_ID!=0
and /*NO LOAD BALANCE*/.
Toshihiro Kitagawa [Mon, 7 Mar 2011 07:00:14 +0000 (07:00 +0000)]
Change REAL_MASTER_NODE_ID to consider PRIMARY_NODE_ID.
Toshihiro Kitagawa [Mon, 7 Mar 2011 06:48:26 +0000 (06:48 +0000)]
Fix bug with load balancing such as following when PRIMARY_NODE_ID is
not same as REAL_MASTER_NODE_ID.
testdb=> BEGIN;
BEGIN
testdb=> SELECT * FROM t1;
ERROR: relation "t1" does not exist
LINE 1: SELECT * FROM t1;
^
testdb=> SELECT * FROM t2;
id
----
(0 rows)
testdb=> COMMIT;
ROLLBACK
Toshihiro Kitagawa [Mon, 7 Mar 2011 06:35:27 +0000 (06:35 +0000)]
Fix insert_lock() so that it extracts sequence name directly.
Toshihiro Kitagawa [Mon, 7 Mar 2011 05:56:58 +0000 (05:56 +0000)]
Change session information to integrate Portal and PreparedStatement
structure into POOL_SENT_MESSAGE structure.
Portal and PreparedStatement structure were used by only extended query
protocol before. but POOL_SENT_MESSAGE is used by simple query protocol too.
This change fixes issues that happen when two protocols were mixed in
one session.
PHP example)
// PREPARE using simple query protocol
pg_query($dbconn, 'PREPARE my_query(TEXT) AS SELECT $1::TEXT');
// EXECUTE using exetended query protocol
pg_execute($dbconn, "my_query", array('test'));
Tatsuo Ishii [Mon, 28 Feb 2011 00:50:51 +0000 (00:50 +0000)]
Fix pgpool-regclass() to be compiled in PostgreSQL 8.0 or later.
7.4 still produces errors but I doubt fixing effort is worth the trouble.
Tatsuo Ishii [Wed, 23 Feb 2011 07:21:19 +0000 (07:21 +0000)]
Update TODO.
Tatsuo Ishii [Wed, 23 Feb 2011 06:47:48 +0000 (06:47 +0000)]
Do not keep persistent connection to backends.
The connection may become bogus if a node down and then wake up
between replication delay checking period.
Restarting worker process is not feasible since there is a time lang
between backend down and restarting worker process.
Tatsuo Ishii [Wed, 23 Feb 2011 06:45:44 +0000 (06:45 +0000)]
Add error message to send_simplequery_message().
Enhance do_query() so that it emmits an error message when receiving ERROR
from backend.
Tatsuo Ishii [Wed, 23 Feb 2011 03:50:01 +0000 (03:50 +0000)]
Fix typo in comment.