10BC0 Support Linux's oom_score_adj API as well as the older oom_adj API. · postgres/postgres@93f4d7f · GitHub
[go: up one dir, main page]
Skip to content

Commit 93f4d7f

Browse files
tglsfdctglsfdc
committed
Support Linux's oom_score_adj API as well as the older oom_adj API.
The simplest way to handle this is just to copy-and-paste the relevant code block in fork_process.c, so that's what I did. (It's possible that something more complicated would be useful to packagers who want to work with either the old or the new API; but at this point the number of such people is rapidly approaching zero, so let's just get the minimal thing done.) Update relevant documentation as well.
1 parent b9212e3 commit 93f4d7f

File tree

3 files changed

+59
-16
lines changed

3 files changed

+59
-16
lines changed

‎contrib/start-scripts/linux‎

Lines changed: 10 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -42,10 +42,14 @@ PGLOG="$PGDATA/serverlog"
4242

4343
# It's often a good idea to protect the postmaster from being killed by the
4444
# OOM killer (which will tend to preferentially kill the postmaster because
45-
# of the way it accounts for shared memory). Setting the OOM_ADJ value to
46-
# -17 will disable OOM kill altogether. If you enable this, you probably want
47-
# to compile PostgreSQL with "-DLINUX_OOM_ADJ=0", so that individual backends
48-
# can still be killed by the OOM killer.
45+
# of the way it accounts for shared memory). Setting the OOM_SCORE_ADJ value
46+
# to -1000 will disable OOM kill altogether. If you enable this, you probably
47+
# want to compile PostgreSQL with "-DLINUX_OOM_SCORE_ADJ=0", so that
48+
# individual backends can still be killed by the OOM killer.
49+
#OOM_SCORE_ADJ=-1000
50+
# Older Linux kernels may not have /proc/self/oom_score_adj, but instead
51+
# /proc/self/oom_adj, which works similarly except the disable value is -17.
52+
# For such a system, enable this and compile with "-DLINUX_OOM_ADJ=0".
4953
#OOM_ADJ=-17
5054

5155
## STOP EDITING HERE
@@ -78,6 +82,7 @@ test -x $DAEMON ||
7882
case $1 in
7983
start)
8084
echo -n "Starting PostgreSQL: "
85+
test x"$OOM_SCORE_ADJ" != x && echo "$OOM_SCORE_ADJ" > /proc/self/oom_score_adj
8186
test x"$OOM_ADJ" != x && echo "$OOM_ADJ" > /proc/self/oom_adj
8287
su - $PGUSER -c "$DAEMON -D '$PGDATA' &" >>$PGLOG 2>&1
8388
echo "ok"
@@ -90,6 +95,7 @@ case $1 in
9095
restart)
9196
echo -n "Restarting PostgreSQL: "
9297
su - $PGUSER -c "$PGCTL stop -D '$PGDATA' -s -m fast -w"
98+
test x"$OOM_SCORE_ADJ" != x && echo "$OOM_SCORE_ADJ" > /proc/self/oom_score_adj
9399
test x"$OOM_ADJ" != x && echo "$OOM_ADJ" > /proc/self/oom_adj
94100
su - $PGUSER -c "$DAEMON -D '$PGDATA' &" >>$PGLOG 2>&1
95101
echo "ok"

‎doc/src/sgml/runtime.sgml‎

Lines changed: 16 additions & 7 deletions
Original file line numberDiff line numberDiff line change
@@ -1268,7 +1268,7 @@ default:\
12681268
In Linux 2.4 and later, the default virtual memory behavior is not
12691269
optimal for <productname>PostgreSQL</productname>. Because of the
12701270
way that the kernel implements memory overcommit, the kernel might
1271-
terminate the <productname>PostgreSQL</productname> server (the
1271+
terminate the <productname>PostgreSQL</productname> postmaster (the
12721272
master server process) if the memory demands of
12731273
another process cause the system to run out of virtual memory.
12741274
</para>
@@ -1317,22 +1317,31 @@ sysctl -w vm.overcommit_memory=2
13171317
<para>
13181318
Another approach, which can be used with or without altering
13191319
<varname>vm.overcommit_memory</>, is to set the process-specific
1320-
<varname>oom_adj</> value for the postmaster process to <literal>-17</>,
1321-
thereby guaranteeing it will not be targeted by the OOM killer. The
1322-
simplest way to do this is to execute
1320+
<varname>oom_score_adj</> value for the postmaster process to
1321+
<literal>-1000</>, thereby guaranteeing it will not be targeted by the OOM
1322+
killer. The simplest way to do this is to execute
13231323
<programlisting>
1324-
echo -17 > /proc/self/oom_adj
1324+
echo -1000 > /proc/self/oom_score_adj
13251325
</programlisting>
13261326
in the postmaster's startup script just before invoking the postmaster.
13271327
Note that this action must be done as root, or it will have no effect;
13281328
so a root-owned startup script is the easiest place to do it. If you
13291329
do this, you may also wish to build <productname>PostgreSQL</>
1330-
with <literal>-DLINUX_OOM_ADJ=0</> added to <varname>CPPFLAGS</>.
1330+
with <literal>-DLINUX_OOM_SCORE_ADJ=0</> added to <varname>CPPFLAGS</>.
13311331
That will cause postmaster child processes to run with the normal
1332-
<varname>oom_adj</> value of zero, so that the OOM killer can still
1332+
<varname>oom_score_adj</> value of zero, so that the OOM killer can still
13331333
target them at need.
13341334
</para>
13351335

1336+
<para>
1337+
Older Linux kernels do not offer <filename>/proc/self/oom_score_adj</>,
1338+
but may have a previous version of the same functionality called
1339+
<filename>/proc/self/oom_adj</>. This works the same except the disable
1340+
value is <literal>-17</> not <literal>-1000</>. The corresponding
1341+
build flag for <productname>PostgreSQL</> is
1342+
<literal>-DLINUX_OOM_ADJ=0</>.
1343+
</para>
1344+
13361345
<note>
13371346
<para>
13381347
Some vendors' Linux 2.4 kernels are reported to have early versions

‎src/backend/postmaster/fork_process.c‎

Lines changed: 33 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -68,12 +68,40 @@ fork_process(void)
6868
* process sizes *including shared memory*. (This is unbelievably
6969
* stupid, but the kernel hackers seem uninterested in improving it.)
7070
* Therefore it's often a good idea to protect the postmaster by
71-
* setting its oom_adj value negative (which has to be done in a
72-
* root-owned startup script). If you just do that much, all child
71+
* setting its oom_score_adj value negative (which has to be done in a
72+
* root-owned startup script). If you just do that much, all child
7373
* processes will also be protected against OOM kill, which might not
74-
* be desirable. You can then choose to build with LINUX_OOM_ADJ
75-
* #defined to 0, or some other value that you want child processes to
76-
* adopt here.
74+
* be desirable. You can then choose to build with
75+
* LINUX_OOM_SCORE_ADJ #defined to 0, or to some other value that you
76+
* want child processes to adopt here.
77+
*/
78+
#ifdef LINUX_OOM_SCORE_ADJ
79+
{
80+
/*
81+
* Use open() not stdio, to ensure we control the open flags. Some
82+
* Linux security environments reject anything but O_WRONLY.
83+
*/
84+
int fd = open("/proc/self/oom_score_adj", O_WRONLY, 0);
85+
86+
/* We ignore all errors */
87+
if (fd >= 0)
88+
{
89+
char buf[16];
90+
int rc;
91+
92+
snprintf(buf, sizeof(buf), "%d\n", LINUX_OOM_SCORE_ADJ);
93+
rc = write(fd, buf, strlen(buf));
94+
(void) rc;
95+
close(fd);
96+
}
97+
}
98+
#endif /* LINUX_OOM_SCORE_ADJ */
99+
100+
/*
101+
* Older Linux kernels have oom_adj not oom_score_adj. This works
102+
* similarly except with a different scale of adjustment values.
103+
* If it's necessary to build Postgres to work with either API,
104+
* you can define both LINUX_OOM_SCORE_ADJ and LINUX_OOM_ADJ.
77105
*/
78106
#ifdef LINUX_OOM_ADJ
79107
{

0 commit comments

Comments
 (0)
0