Hi,
2018-09-14 12:23 GMT+02:00 Masahiko Sawada <[email protected]>:
>> 2. If we know that this is neither superuser nor replication connection, we
>> should check that there are at least (superuser_reserved_connections +
>> NumWalSenders() - max_wal_senders) connection slots are available.
>
> You wanted to mean (superuser_reserved_connections + max_wal_senders -
> NumWalSenders()) in the second point?
Sure, my bad. Did a mistake when writing an email, but in the attached
file it looks good.
>
> One argrable point of the second option could be that it breaks
> backward compatibility of the parameter configurations. That is, the
> existing systems need to re-configure the max_connections. So it might
> be better to take the first option with
> replication_reservd_connections = 0 by default.
Please find attached the new version of the patch, which introduces
replication_reservd_connections GUC
Regards,
--
Alexander Kukushkin
diff --git a/doc/src/sgml/config.sgml b/doc/src/sgml/config.sgml
index e1073ac6d3..80e6ef9f67 100644
--- a/doc/src/sgml/config.sgml
+++ b/doc/src/sgml/config.sgml
@@ -3059,6 +3059,32 @@ include_dir 'conf.d'
</listitem>
</varlistentry>
+ <varlistentry id="guc-replication-reserved-connections"
+ xreflabel="replication_reserved_connections">
+ <term><varname>replication_reserved_connections</varname>
+ (<type>integer</type>)
+ <indexterm>
+ <primary><varname>replication_reserved_connections</varname> configuration parameter</primary>
+ </indexterm>
+ </term>
+ <listitem>
+ <para>
+ Determines the number of connection <quote>slots</quote> that
+ are reserved for replication connections. Whenever the number
+ of active concurrent connections is at least
+ <varname>max_connections</varname> minus
+ <varname>replication_reserved_connections</varname> plus
+ <literal>number of active wal senders</literal>, new
+ non-superuser and non-replication connections will not be accepted.
+ </para>
+
+ <para>
+ The default value is zero. The value should not exceed <varname>max_wal_senders</varname>.
+ This parameter can only be set at server start.
+ </para>
+ </listitem>
+ </varlistentry>
+
<varlistentry id="guc-max-replication-slots" xreflabel="max_replication_slots">
<term><varname>max_replication_slots</varname> (<type>integer</type>)
<indexterm>
diff --git a/src/backend/postmaster/postmaster.c b/src/backend/postmaster/postmaster.c
index 305ff36258..a5a95ee92c 100644
--- a/src/backend/postmaster/postmaster.c
+++ b/src/backend/postmaster/postmaster.c
@@ -903,6 +903,10 @@ PostmasterMain(int argc, char *argv[])
if (max_wal_senders > 0 && wal_level == WAL_LEVEL_MINIMAL)
ereport(ERROR,
(errmsg("WAL streaming (max_wal_senders > 0) requires wal_level \"replica\" or \"logical\"")));
+ if (replication_reserved_connections > max_wal_senders)
+ ereport(WARNING,
+ (errmsg("Value of replication_reserved_connections (%d) exceeds value of max_wal_senders (%d)",
+ replication_reserved_connections, max_wal_senders)));
/*
* Other one-time internal sanity checks can go here, if they are fast.
diff --git a/src/backend/replication/walsender.c b/src/backend/replication/walsender.c
index 370429d746..e64d5ed44d 100644
--- a/src/backend/replication/walsender.c
+++ b/src/backend/replication/walsender.c
@@ -122,6 +122,8 @@ int max_wal_senders = 0; /* the maximum number of concurrent
int wal_sender_timeout = 60 * 1000; /* maximum time to send one WAL
* data message */
bool log_replication_commands = false;
+int replication_reserved_connections = 0; /* the number of connection slots
+ * reserved for replication connections */
/*
* State for WalSndWakeupRequest
@@ -2284,6 +2286,10 @@ InitWalSenderSlot(void)
walsnd->applyLag = -1;
walsnd->state = WALSNDSTATE_STARTUP;
walsnd->latch = &MyProc->procLatch;
+
+ /* increment the number of allocated wal sender slots */
+ pg_atomic_fetch_add_u32(&WalSndCtl->num_wal_senders, 1);
+
SpinLockRelease(&walsnd->mutex);
/* don't need the lock anymore */
MyWalSnd = (WalSnd *) walsnd;
@@ -2317,6 +2323,10 @@ WalSndKill(int code, Datum arg)
walsnd->latch = NULL;
/* Mark WalSnd struct as no longer being in use. */
walsnd->pid = 0;
+
+ /* decrement the number of allocated wal sender slots */
+ pg_atomic_fetch_sub_u32(&WalSndCtl->num_wal_senders, 1);
+
SpinLockRelease(&walsnd->mutex);
}
@@ -3033,6 +3043,7 @@ WalSndShmemInit(void)
{
/* First time through, so initialize */
MemSet(WalSndCtl, 0, WalSndShmemSize());
+ pg_atomic_init_u32(&WalSndCtl->num_wal_senders, 0);
for (i = 0; i < NUM_SYNC_REP_WAIT_MODE; i++)
SHMQueueInit(&(WalSndCtl->SyncRepQueue[i]));
@@ -3587,3 +3598,9 @@ LagTrackerRead(int head, XLogRecPtr lsn, TimestampTz now)
Assert(time != 0);
return now - time;
}
+
+/* Return the amount of allocated wal_sender slots */
+uint32 NumWalSenders(void)
+{
+ return pg_atomic_read_u32(&WalSndCtl->num_wal_senders);
+}
diff --git a/src/backend/utils/init/postinit.c b/src/backend/utils/init/postinit.c
index 5ef6315d20..436574e85d 100644
--- a/src/backend/utils/init/postinit.c
+++ b/src/backend/utils/init/postinit.c
@@ -789,17 +789,28 @@ InitPostgres(const char *in_dbname, Oid dboid, const char *username,
}
/*
- * The last few connection slots are reserved for superusers. Although
- * replication connections currently require superuser privileges, we
- * don't allow them to consume the reserved slots, which are intended for
- * interactive use.
+ * The last few connection slots are reserved for superusers and
+ * replication. Superusers always have a priority over replication
+ * connections.
*/
- if ((!am_superuser || am_walsender) &&
- ReservedBackends > 0 &&
- !HaveNFreeProcs(ReservedBackends))
- ereport(FATAL,
- (errcode(ERRCODE_TOO_MANY_CONNECTIONS),
- errmsg("remaining connection slots are reserved for non-replication superuser connections")));
+ if (am_walsender)
+ {
+ if (ReservedBackends > 0 && !HaveNFreeProcs(ReservedBackends))
+ ereport(FATAL,
+ (errcode(ERRCODE_TOO_MANY_CONNECTIONS),
+ errmsg("remaining connection slots are reserved for non-replication superuser connections")));
+ }
+ else if (!am_superuser)
+ {
+ uint32 n = Min(max_wal_senders, replication_reserved_connections);
+
+ n = ReservedBackends + n - NumWalSenders();
+
+ if (n > 0 && !HaveNFreeProcs(n))
+ ereport(FATAL,
+ (errcode(ERRCODE_TOO_MANY_CONNECTIONS),
+ errmsg("remaining connection slots are reserved for replication or superuser connections")));
+ }
/* Check replication permissions needed for walsender processes. */
if (am_walsender)
diff --git a/src/backend/utils/misc/guc.c b/src/backend/utils/misc/guc.c
index 77662aff7f..5203b70ef6 100644
--- a/src/backend/utils/misc/guc.c
+++ b/src/backend/utils/misc/guc.c
@@ -2527,6 +2527,17 @@ static struct config_int ConfigureNamesInt[] =
NULL, NULL, NULL
},
+ {
+ /* see max_connections, max_wal_senders and superuser_reserved_connections */
+ {"replication_reserved_connections", PGC_POSTMASTER, CONN_AUTH_SETTINGS,
+ gettext_noop("Sets the number of connection slots reserved for replication connections."),
+ NULL
+ },
+ &replication_reserved_connections,
+ 0, 0, MAX_BACKENDS,
+ NULL, NULL, NULL
+ },
+
{
/* see max_wal_senders */
{"max_replication_slots", PGC_POSTMASTER, REPLICATION_SENDING,
diff --git a/src/backend/utils/misc/postgresql.conf.sample b/src/backend/utils/misc/postgresql.conf.sample
index 4e61bc6521..6713833c5b 100644
--- a/src/backend/utils/misc/postgresql.conf.sample
+++ b/src/backend/utils/misc/postgresql.conf.sample
@@ -237,6 +237,8 @@
#max_wal_senders = 10 # max number of walsender processes
# (change requires restart)
+#replication_reserved_connections = 0 # number of connection slots reserved
+ # for replication connections. (change requires restart)
#wal_keep_segments = 0 # in logfile segments; 0 disables
#wal_sender_timeout = 60s # in milliseconds; 0 disables
diff --git a/src/include/replication/walsender.h b/src/include/replication/walsender.h
index 45b72a76db..d91f9348ba 100644
--- a/src/include/replication/walsender.h
+++ b/src/include/replication/walsender.h
@@ -36,7 +36,9 @@ extern bool wake_wal_senders;
extern int max_wal_senders;
extern int wal_sender_timeout;
extern bool log_replication_commands;
+extern int replication_reserved_connections;
+extern uint32 NumWalSenders(void);
extern void InitWalSender(void);
extern bool exec_replication_command(const char *query_string);
extern void WalSndErrorCleanup(void);
diff --git a/src/include/replication/walsender_private.h b/src/include/replication/walsender_private.h
index 4b90477936..3ef8f4d7b8 100644
--- a/src/include/replication/walsender_private.h
+++ b/src/include/replication/walsender_private.h
@@ -14,6 +14,7 @@
#include "access/xlog.h"
#include "nodes/nodes.h"
+#include "port/atomics.h"
#include "replication/syncrep.h"
#include "storage/latch.h"
#include "storage/shmem.h"
@@ -101,6 +102,8 @@ typedef struct
*/
bool sync_standbys_defined;
+ pg_atomic_uint32 num_wal_senders;
+
WalSnd walsnds[FLEXIBLE_ARRAY_MEMBER];
} WalSndCtlData;
