Movatterモバイル変換

[0]ホーム
URL:

Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

 Sign up
Appearance settings

Commitc30f54a

Browse files
committed
Detect POLLHUP/POLLRDHUP while running queries.
Provide a new GUC check_client_connection_interval that can be used tocheck whether the client connection has gone away, while running verylong queries. It is disabled by default.For now this uses a non-standard Linux extension (also adopted by atleast one other OS). POLLRDHUP is not defined by POSIX, and other OSesdon't have a reliable way to know if a connection was closed withoutactually trying to read or write.In future we might consider trying to send a no-op/heartbeat messageinstead, but that could require protocol changes.Author: Sergey Cherkashin <s.cherkashin@postgrespro.ru>Author: Thomas Munro <thomas.munro@gmail.com>Reviewed-by: Thomas Munro <thomas.munro@gmail.com>Reviewed-by: Tatsuo Ishii <ishii@sraoss.co.jp>Reviewed-by: Konstantin Knizhnik <k.knizhnik@postgrespro.ru>Reviewed-by: Zhihong Yu <zyu@yugabyte.com>Reviewed-by: Andres Freund <andres@anarazel.de>Reviewed-by: Maksim Milyutin <milyutinma@gmail.com>Reviewed-by: Tsunakawa, Takayuki/綱川 貴之 <tsunakawa.takay@fujitsu.com>Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us> (much earlier version)Discussion:https://postgr.es/m/77def86b27e41f0efcba411460e929ae%40postgrespro.ru
1 parent174edbe commitc30f54a

File tree

11 files changed

+156
-0
lines changed

11 files changed

+156
-0
lines changed

‎doc/src/sgml/config.sgml

Lines changed: 37 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -998,6 +998,43 @@ include_dir 'conf.d'
998998
</listitem>
999999
</varlistentry>
10001000

1001+
<varlistentry id="guc-client-connection-check-interval" xreflabel="client_connection_check_interval">
1002+
<term><varname>client_connection_check_interval</varname> (<type>integer</type>)
1003+
<indexterm>
1004+
<primary><varname>client_connection_check_interval</varname> configuration parameter</primary>
1005+
</indexterm>
1006+
</term>
1007+
<listitem>
1008+
<para>
1009+
Sets the time interval between optional checks that the client is still
1010+
connected, while running queries. The check is performed by polling
1011+
the socket, and allows long running queries to be aborted sooner if
1012+
the kernel reports that the connection is closed.
1013+
</para>
1014+
<para>
1015+
This option is currently available only on systems that support the
1016+
non-standard <symbol>POLLRDHUP</symbol> extension to the
1017+
<symbol>poll</symbol> system call, including Linux.
1018+
</para>
1019+
<para>
1020+
If the value is specified without units, it is taken as milliseconds.
1021+
The default value is <literal>0</literal>, which disables connection
1022+
checks. Without connection checks, the server will detect the loss of
1023+
the connection only at the next interaction with the socket, when it
1024+
waits for, receives or sends data.
1025+
</para>
1026+
<para>
1027+
For the kernel itself to detect lost TCP connections reliably and within
1028+
a known timeframe in all scenarios including network failure, it may
1029+
also be necessary to adjust the TCP keepalive settings of the operating
1030+
system, or the <xref linkend="guc-tcp-keepalives-idle"/>,
1031+
<xref linkend="guc-tcp-keepalives-interval"/> and
1032+
<xref linkend="guc-tcp-keepalives-count"/> settings of
1033+
<productname>PostgreSQL</productname>.
1034+
</para>
1035+
</listitem>
1036+
</varlistentry>
1037+
10011038
</variablelist>
10021039
</sect2>
10031040

‎src/backend/libpq/pqcomm.c

Lines changed: 40 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -54,6 +54,9 @@
5454
*/
5555
#include"postgres.h"
5656

57+
#ifdefHAVE_POLL_H
58+
#include<poll.h>
59+
#endif
5760
#include<signal.h>
5861
#include<fcntl.h>
5962
#include<grp.h>
@@ -1921,3 +1924,40 @@ pq_settcpusertimeout(int timeout, Port *port)
19211924

19221925
returnSTATUS_OK;
19231926
}
1927+
1928+
/*
1929+
* Check if the client is still connected.
1930+
*/
1931+
bool
1932+
pq_check_connection(void)
1933+
{
1934+
#if defined(POLLRDHUP)
1935+
/*
1936+
* POLLRDHUP is a Linux extension to poll(2) to detect sockets closed by
1937+
* the other end. We don't have a portable way to do that without
1938+
* actually trying to read or write data on other systems. We don't want
1939+
* to read because that would be confused by pipelined queries and COPY
1940+
* data. Perhaps in future we'll try to write a heartbeat message instead.
1941+
*/
1942+
structpollfdpollfd;
1943+
intrc;
1944+
1945+
pollfd.fd=MyProcPort->sock;
1946+
pollfd.events=POLLOUT |POLLIN |POLLRDHUP;
1947+
pollfd.revents=0;
1948+
1949+
rc=poll(&pollfd,1,0);
1950+
1951+
if (rc<0)
1952+
{
1953+
ereport(COMMERROR,
1954+
(errcode_for_socket_access(),
1955+
errmsg("could not poll socket: %m")));
1956+
return false;
1957+
}
1958+
elseif (rc==1&& (pollfd.revents& (POLLHUP |POLLRDHUP)))
1959+
return false;
1960+
#endif
1961+
1962+
return true;
1963+
}

‎src/backend/tcop/postgres.c

Lines changed: 32 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -102,6 +102,9 @@ intmax_stack_depth = 100;
102102
/* wait N seconds to allow attach from a debugger */
103103
intPostAuthDelay=0;
104104

105+
/* Time between checks that the client is still connected. */
106+
intclient_connection_check_interval=0;
107+
105108
/* ----------------
106109
*private typedefs etc
107110
* ----------------
@@ -2671,6 +2674,14 @@ start_xact_command(void)
26712674
* not desired, the timeout has to be disabled explicitly.
26722675
*/
26732676
enable_statement_timeout();
2677+
2678+
/* Start timeout for checking if the client has gone away if necessary. */
2679+
if (client_connection_check_interval>0&&
2680+
IsUnderPostmaster&&
2681+
MyProcPort&&
2682+
!get_timeout_active(CLIENT_CONNECTION_CHECK_TIMEOUT))
2683+
enable_timeout_after(CLIENT_CONNECTION_CHECK_TIMEOUT,
2684+
client_connection_check_interval);
26742685
}
26752686

26762687
staticvoid
@@ -3149,6 +3160,27 @@ ProcessInterrupts(void)
31493160
(errcode(ERRCODE_ADMIN_SHUTDOWN),
31503161
errmsg("terminating connection due to administrator command")));
31513162
}
3163+
3164+
if (CheckClientConnectionPending)
3165+
{
3166+
CheckClientConnectionPending= false;
3167+
3168+
/*
3169+
* Check for lost connection and re-arm, if still configured, but not
3170+
* if we've arrived back at DoingCommandRead state. We don't want to
3171+
* wake up idle sessions, and they already know how to detect lost
3172+
* connections.
3173+
*/
3174+
if (!DoingCommandRead&&client_connection_check_interval>0)
3175+
{
3176+
if (!pq_check_connection())
3177+
ClientConnectionLost= true;
3178+
else
3179+
enable_timeout_after(CLIENT_CONNECTION_CHECK_TIMEOUT,
3180+
client_connection_check_interval);
3181+
}
3182+
}
3183+
31523184
if (ClientConnectionLost)
31533185
{
31543186
QueryCancelPending= false;/* lost connection trumps QueryCancel */

‎src/backend/utils/init/globals.c

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -30,6 +30,7 @@ ProtocolVersion FrontendProtocol;
3030
volatilesig_atomic_tInterruptPending= false;
3131
volatilesig_atomic_tQueryCancelPending= false;
3232
volatilesig_atomic_tProcDiePending= false;
33+
volatilesig_atomic_tCheckClientConnectionPending= false;
3334
volatilesig_atomic_tClientConnectionLost= false;
3435
volatilesig_atomic_tIdleInTransactionSessionTimeoutPending= false;
3536
volatilesig_atomic_tIdleSessionTimeoutPending= false;

‎src/backend/utils/init/postinit.c

Lines changed: 10 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -73,6 +73,7 @@ static void StatementTimeoutHandler(void);
7373
staticvoidLockTimeoutHandler(void);
7474
staticvoidIdleInTransactionSessionTimeoutHandler(void);
7575
staticvoidIdleSessionTimeoutHandler(void);
76+
staticvoidClientCheckTimeoutHandler(void);
7677
staticboolThereIsAtLeastOneRole(void);
7778
staticvoidprocess_startup_options(Port*port,boolam_superuser);
7879
staticvoidprocess_settings(Oiddatabaseid,Oidroleid);
@@ -620,6 +621,7 @@ InitPostgres(const char *in_dbname, Oid dboid, const char *username,
620621
RegisterTimeout(IDLE_IN_TRANSACTION_SESSION_TIMEOUT,
621622
IdleInTransactionSessionTimeoutHandler);
622623
RegisterTimeout(IDLE_SESSION_TIMEOUT,IdleSessionTimeoutHandler);
624+
RegisterTimeout(CLIENT_CONNECTION_CHECK_TIMEOUT,ClientCheckTimeoutHandler);
623625
}
624626

625627
/*
@@ -1242,6 +1244,14 @@ IdleSessionTimeoutHandler(void)
12421244
SetLatch(MyLatch);
12431245
}
12441246

1247+
staticvoid
1248+
ClientCheckTimeoutHandler(void)
1249+
{
1250+
CheckClientConnectionPending= true;
1251+
InterruptPending= true;
1252+
SetLatch(MyLatch);
1253+
}
1254+
12451255
/*
12461256
* Returns true if at least one role is defined in this database cluster.
12471257
*/

‎src/backend/utils/misc/guc.c

Lines changed: 29 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -20,6 +20,9 @@
2020
#include<float.h>
2121
#include<math.h>
2222
#include<limits.h>
23+
#ifdefHAVE_POLL_H
24+
#include<poll.h>
25+
#endif
2326
#ifndefWIN32
2427
#include<sys/mman.h>
2528
#endif
@@ -204,6 +207,7 @@ static bool check_autovacuum_work_mem(int *newval, void **extra, GucSource sourc
204207
staticboolcheck_effective_io_concurrency(int*newval,void**extra,GucSourcesource);
205208
staticboolcheck_maintenance_io_concurrency(int*newval,void**extra,GucSourcesource);
206209
staticboolcheck_huge_page_size(int*newval,void**extra,GucSourcesource);
210+
staticboolcheck_client_connection_check_interval(int*newval,void**extra,GucSourcesource);
207211
staticvoidassign_pgstat_temp_directory(constchar*newval,void*extra);
208212
staticboolcheck_application_name(char**newval,void**extra,GucSourcesource);
209213
staticvoidassign_application_name(constchar*newval,void*extra);
@@ -3501,6 +3505,17 @@ static struct config_int ConfigureNamesInt[] =
35013505
NULL,NULL,NULL
35023506
},
35033507

3508+
{
3509+
{"client_connection_check_interval",PGC_USERSET,CLIENT_CONN_OTHER,
3510+
gettext_noop("Sets the time interval between checks for disconnection while running queries."),
3511+
NULL,
3512+
GUC_UNIT_MS
3513+
},
3514+
&client_connection_check_interval,
3515+
0,0,INT_MAX,
3516+
check_client_connection_check_interval,NULL,NULL
3517+
},
3518+
35043519
/* End-of-list marker */
35053520
{
35063521
{NULL,0,0,NULL,NULL},NULL,0,0,0,NULL,NULL,NULL
@@ -11980,6 +11995,20 @@ check_huge_page_size(int *newval, void **extra, GucSource source)
1198011995
return true;
1198111996
}
1198211997

11998+
staticbool
11999+
check_client_connection_check_interval(int*newval,void**extra,GucSourcesource)
12000+
{
12001+
#ifndefPOLLRDHUP
12002+
/* Linux only, for now. See pq_check_connection(). */
12003+
if (*newval!=0)
12004+
{
12005+
GUC_check_errdetail("client_connection_check_interval must be set to 0 on platforms that lack POLLRDHUP.");
12006+
return false;
12007+
}
12008+
#endif
12009+
return true;
12010+
}
12011+
1198312012
staticvoid
1198412013
assign_pgstat_temp_directory(constchar*newval,void*extra)
1198512014
{

‎src/backend/utils/misc/postgresql.conf.sample

Lines changed: 3 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -719,6 +719,9 @@
719719

720720
#dynamic_library_path = '$libdir'
721721

722+
#client_connection_check_interval = 0# time between checks for client
723+
# disconnection while running queries;
724+
# 0 for never
722725

723726
#------------------------------------------------------------------------------
724727
# LOCK MANAGEMENT

‎src/include/libpq/libpq.h

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -71,6 +71,7 @@ extern intpq_getbyte(void);
7171
externintpq_peekbyte(void);
7272
externintpq_getbyte_if_available(unsignedchar*c);
7373
externintpq_putmessage_v2(charmsgtype,constchar*s,size_tlen);
74+
externboolpq_check_connection(void);
7475

7576
/*
7677
* prototypes for functions in be-secure.c

‎src/include/miscadmin.h

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -85,6 +85,7 @@ extern PGDLLIMPORT volatile sig_atomic_t IdleInTransactionSessionTimeoutPending;
8585
externPGDLLIMPORTvolatilesig_atomic_tIdleSessionTimeoutPending;
8686
externPGDLLIMPORTvolatilesig_atomic_tProcSignalBarrierPending;
8787

88+
externPGDLLIMPORTvolatilesig_atomic_tCheckClientConnectionPending;
8889
externPGDLLIMPORTvolatilesig_atomic_tClientConnectionLost;
8990

9091
/* these are marked volatile because they are examined by signal handlers: */

‎src/include/tcop/tcopprot.h

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -29,6 +29,7 @@ extern CommandDest whereToSendOutput;
2929
externPGDLLIMPORTconstchar*debug_query_string;
3030
externintmax_stack_depth;
3131
externintPostAuthDelay;
32+
externintclient_connection_check_interval;
3233

3334
/* GUC-configurable parameters */
3435

‎src/include/utils/timeout.h

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -32,6 +32,7 @@ typedef enum TimeoutId
3232
STANDBY_LOCK_TIMEOUT,
3333
IDLE_IN_TRANSACTION_SESSION_TIMEOUT,
3434
IDLE_SESSION_TIMEOUT,
35+
CLIENT_CONNECTION_CHECK_TIMEOUT,
3536
/* First user-definable timeout reason */
3637
USER_TIMEOUT,
3738
/* Maximum number of timeout reasons */

0 commit comments

Comments
 (0)
[8]ページ先頭
©2009-2025 Movatter.jp