Improve our mechanism for controlling the Linux out-of-memory killer.

Arrange for postmaster child processes to respond to two environment
variables, PG_OOM_ADJUST_FILE and PG_OOM_ADJUST_VALUE, to determine whether
they reset their OOM score adjustments and if so to what.  This is superior
to the previous design involving #ifdef's in several ways.  The behavior is
now available in a default build, and both ends of the adjustment --- the
original adjustment of the postmaster's level and the subsequent
readjustment by child processes --- can now be controlled in one place,
namely the postmaster launch script.  So it's no longer necessary for the
launch script to act on faith that the server was compiled with the
appropriate options.  In addition, if someone wants to use an OOM score
other than zero for the child processes, that doesn't take a recompile
anymore; and we no longer have to cater separately to the two different
historical kernel APIs for this adjustment.

Gurjeet Singh, somewhat revised by me

Author: tglsfdc

contrib/start-scripts/linux

@@ -43,14 +43,17 @@ PGLOG="$PGDATA/serverlog"
 # It's often a good idea to protect the postmaster from being killed by the
 # OOM killer (which will tend to preferentially kill the postmaster because
 # of the way it accounts for shared memory).  Setting the OOM_SCORE_ADJ value
-# to -1000 will disable OOM kill altogether.  If you enable this, you probably
-# want to compile PostgreSQL with "-DLINUX_OOM_SCORE_ADJ=0", so that
-# individual backends can still be killed by the OOM killer.
+# to -1000 will disable OOM kill altogether, which is a good thing for the
+# postmaster, but not so much for individual backends.  If you enable this,
+# also uncomment the DAEMON_ENV line, which will instruct backends to set
+# their OOM adjustments back to the default setting of zero.
 #OOM_SCORE_ADJ=-1000
+#DAEMON_ENV="PG_OOM_ADJUST_FILE=/proc/self/oom_score_adj"
 # Older Linux kernels may not have /proc/self/oom_score_adj, but instead
 # /proc/self/oom_adj, which works similarly except the disable value is -17.
-# For such a system, enable this and compile with "-DLINUX_OOM_ADJ=0".
+# For such a system, uncomment these two lines instead.
 #OOM_ADJ=-17
+#DAEMON_ENV="PG_OOM_ADJUST_FILE=/proc/self/oom_adj"
 
 ## STOP EDITING HERE
 
@@ -84,7 +87,7 @@ case $1 in
 	echo -n "Starting PostgreSQL: "
 	test x"$OOM_SCORE_ADJ" != x && echo "$OOM_SCORE_ADJ" > /proc/self/oom_score_adj
 	test x"$OOM_ADJ" != x && echo "$OOM_ADJ" > /proc/self/oom_adj
-	su - $PGUSER -c "$DAEMON -D '$PGDATA' &" >>$PGLOG 2>&1
+	su - $PGUSER -c "$DAEMON_ENV $DAEMON -D '$PGDATA' &" >>$PGLOG 2>&1
 	echo "ok"
 	;;
   stop)
@@ -97,7 +100,7 @@ case $1 in
 	su - $PGUSER -c "$PGCTL stop -D '$PGDATA' -s -m fast -w"
 	test x"$OOM_SCORE_ADJ" != x && echo "$OOM_SCORE_ADJ" > /proc/self/oom_score_adj
 	test x"$OOM_ADJ" != x && echo "$OOM_ADJ" > /proc/self/oom_adj
-	su - $PGUSER -c "$DAEMON -D '$PGDATA' &" >>$PGLOG 2>&1
+	su - $PGUSER -c "$DAEMON_ENV $DAEMON -D '$PGDATA' &" >>$PGLOG 2>&1
 	echo "ok"
 	;;
   reload)

doc/src/sgml/runtime.sgml

@@ -1275,7 +1275,7 @@ sysctl -w vm.overcommit_memory=2
    <para>
     Another approach, which can be used with or without altering
     <varname>vm.overcommit_memory</>, is to set the process-specific
-    <varname>oom_score_adj</> value for the postmaster process to
+    <firstterm>OOM score adjustment</> value for the postmaster process to
     <literal>-1000</>, thereby guaranteeing it will not be targeted by the OOM
     killer.  The simplest way to do this is to execute
 <programlisting>
@@ -1284,20 +1284,28 @@ echo -1000 > /proc/self/oom_score_adj
     in the postmaster's startup script just before invoking the postmaster.
     Note that this action must be done as root, or it will have no effect;
     so a root-owned startup script is the easiest place to do it.  If you
-    do this, you may also wish to build <productname>PostgreSQL</>
-    with <literal>-DLINUX_OOM_SCORE_ADJ=0</> added to <varname>CPPFLAGS</>.
-    That will cause postmaster child processes to run with the normal
-    <varname>oom_score_adj</> value of zero, so that the OOM killer can still
-    target them at need.
+    do this, you should also set these environment variables in the startup
+    script before invoking the postmaster:
+<programlisting>
+export PG_OOM_ADJUST_FILE=/proc/self/oom_score_adj
+export PG_OOM_ADJUST_VALUE=0
+</programlisting>
+    These settings will cause postmaster child processes to run with the
+    normal OOM score adjustment of zero, so that the OOM killer can still
+    target them at need.  You could use some other value for
+    <envar>PG_OOM_ADJUST_VALUE</> if you want the child processes to run
+    with some other OOM score adjustment.  (<envar>PG_OOM_ADJUST_VALUE</>
+    can also be omitted, in which case it defaults to zero.)  If you do not
+    set <envar>PG_OOM_ADJUST_FILE</>, the child processes will run with the
+    same OOM score adjustment as the postmaster, which is unwise since the
+    whole point is to ensure that the postmaster has a preferential setting.
    </para>
 
    <para>
     Older Linux kernels do not offer <filename>/proc/self/oom_score_adj</>,
     but may have a previous version of the same functionality called
     <filename>/proc/self/oom_adj</>.  This works the same except the disable
-    value is <literal>-17</> not <literal>-1000</>.  The corresponding
-    build flag for <productname>PostgreSQL</> is
-    <literal>-DLINUX_OOM_ADJ=0</>.
+    value is <literal>-17</> not <literal>-1000</>.
    </para>
 
    <note>

src/backend/postmaster/fork_process.c

@@ -31,6 +31,7 @@ pid_t
 fork_process(void)
 {
 	pid_t		result;
+	const char *oomfilename;
 
 #ifdef LINUX_PROFILE
 	struct itimerval prof_itimer;
@@ -71,62 +72,40 @@ fork_process(void)
 		 * process sizes *including shared memory*.  (This is unbelievably
 		 * stupid, but the kernel hackers seem uninterested in improving it.)
 		 * Therefore it's often a good idea to protect the postmaster by
-		 * setting its oom_score_adj value negative (which has to be done in a
-		 * root-owned startup script). If you just do that much, all child
-		 * processes will also be protected against OOM kill, which might not
-		 * be desirable.  You can then choose to build with
-		 * LINUX_OOM_SCORE_ADJ #defined to 0, or to some other value that you
-		 * want child processes to adopt here.
+		 * setting its OOM score adjustment negative (which has to be done in
+		 * a root-owned startup script).  Since the adjustment is inherited by
+		 * child processes, this would ordinarily mean that all the
+		 * postmaster's children are equally protected against OOM kill, which
+		 * is not such a good idea.  So we provide this code to allow the
+		 * children to change their OOM score adjustments again.  Both the
+		 * file name to write to and the value to write are controlled by
+		 * environment variables, which can be set by the same startup script
+		 * that did the original adjustment.
 		 */
-#ifdef LINUX_OOM_SCORE_ADJ
-		{
-			/*
-			 * Use open() not stdio, to ensure we control the open flags. Some
-			 * Linux security environments reject anything but O_WRONLY.
-			 */
-			int			fd = open("/proc/self/oom_score_adj", O_WRONLY, 0);
-
-			/* We ignore all errors */
-			if (fd >= 0)
-			{
-				char		buf[16];
-				int			rc;
+		oomfilename = getenv("PG_OOM_ADJUST_FILE");
 
-				snprintf(buf, sizeof(buf), "%d\n", LINUX_OOM_SCORE_ADJ);
-				rc = write(fd, buf, strlen(buf));
-				(void) rc;
-				close(fd);
-			}
-		}
-#endif   /* LINUX_OOM_SCORE_ADJ */
-
-		/*
-		 * Older Linux kernels have oom_adj not oom_score_adj.  This works
-		 * similarly except with a different scale of adjustment values. If
-		 * it's necessary to build Postgres to work with either API, you can
-		 * define both LINUX_OOM_SCORE_ADJ and LINUX_OOM_ADJ.
-		 */
-#ifdef LINUX_OOM_ADJ
+		if (oomfilename != NULL)
 		{
 			/*
 			 * Use open() not stdio, to ensure we control the open flags. Some
 			 * Linux security environments reject anything but O_WRONLY.
 			 */
-			int			fd = open("/proc/self/oom_adj", O_WRONLY, 0);
+			int			fd = open(oomfilename, O_WRONLY, 0);
 
 			/* We ignore all errors */
 			if (fd >= 0)
 			{
-				char		buf[16];
+				const char *oomvalue = getenv("PG_OOM_ADJUST_VALUE");
 				int			rc;
 
-				snprintf(buf, sizeof(buf), "%d\n", LINUX_OOM_ADJ);
-				rc = write(fd, buf, strlen(buf));
+				if (oomvalue == NULL)	/* supply a useful default */
+					oomvalue = "0";
+
+				rc = write(fd, oomvalue, strlen(oomvalue));
 				(void) rc;
 				close(fd);
 			}
 		}
-#endif   /* LINUX_OOM_ADJ */
 
 		/*
 		 * Make sure processes do not share OpenSSL randomness state.