Announcing: ZFS-fuse 0.7.0
#116 — S38zfs-fuse (1142): /proc/1163/oom_adj is deprecated, please use /proc/1163/oom_score_adj instead.
| State | Resolved |
|---|---|
| Version: | 0.6.9 |
| Area | Process |
| Issue type | Patch |
| Severity | Medium |
| Submitted by | (anonymous) |
| Submitted on | Jan 28, 2011 |
| Responsible | Seth Heeren |
| Target release: | 0.7.0 |
Last modified on
Feb 01, 2012
by
Seth Heeren
dmesg reports:
[ 38.447245] cfg80211: Found new beacon on frequency: 2472 MHz (Ch 13) on phy0
[ 38.606987] S38zfs-fuse (1142): /proc/1163/oom_adj is deprecated, please use /proc/1163/oom_score_adj instead.
[ 40.940789] IPv6 over IPv4 tunneling driver
Solution is:
$ diff -u /etc/init.d/zfs-fuse.orig /etc/init.d/zfs-fuse
--- /etc/init.d/zfs-fuse.orig 2011-01-28 21:20:55.725397214 +0100
+++ /etc/init.d/zfs-fuse 2011-01-28 21:21:25.865931918 +0100
@@ -139,7 +139,7 @@
then
log_end_msg 0
log_action_begin_msg "Immunizing $NAME against OOM kills and sendsigs signals"
- echo -17 > /proc/$(cat /var/run/$NAME.pid)/oom_adj
+ echo -17 > /proc/$(cat /var/run/$NAME.pid)/oom_score_adj
ES_TO_REPORT=$?
if [ "$ES_TO_REPORT" = 0 ] ; then
log_action_end_msg 0
[ 38.447245] cfg80211: Found new beacon on frequency: 2472 MHz (Ch 13) on phy0
[ 38.606987] S38zfs-fuse (1142): /proc/1163/oom_adj is deprecated, please use /proc/1163/oom_score_adj instead.
[ 40.940789] IPv6 over IPv4 tunneling driver
Solution is:
$ diff -u /etc/init.d/zfs-fuse.orig /etc/init.d/zfs-fuse
--- /etc/init.d/zfs-fuse.orig 2011-01-28 21:20:55.725397214 +0100
+++ /etc/init.d/zfs-fuse 2011-01-28 21:21:25.865931918 +0100
@@ -139,7 +139,7 @@
then
log_end_msg 0
log_action_begin_msg "Immunizing $NAME against OOM kills and sendsigs signals"
- echo -17 > /proc/$(cat /var/run/$NAME.pid)/oom_adj
+ echo -17 > /proc/$(cat /var/run/$NAME.pid)/oom_score_adj
ES_TO_REPORT=$?
if [ "$ES_TO_REPORT" = 0 ] ; then
log_action_end_msg 0
Added by
(anonymous)
on
Jan 28, 2011 04:27 PM
oops, typo:
--- /etc/init.d/zfs-fuse.orig 2011-01-28 21:20:55.725397214 +0100
+++ /etc/init.d/zfs-fuse 2011-01-28 21:26:32.680999665 +0100
@@ -139,7 +139,7 @@
then
log_end_msg 0
log_action_begin_msg "Immunizing $NAME against OOM kills and sendsigs signals"
- echo -17 > /proc/$(cat /var/run/$NAME.pid)/oom_adj
+ echo -1000 > /proc/$(cat /var/run/$NAME.pid)/oom_score_adj
ES_TO_REPORT=$?
if [ "$ES_TO_REPORT" = 0 ] ; then
log_action_end_msg 0
According to the docs:
The value of /proc/<pid>/oom_score_adj is added to the badness score before it
is used to determine which task to kill. Acceptable values range from -1000
(OOM_SCORE_ADJ_MIN) to +1000 (OOM_SCORE_ADJ_MAX). This allows userspace to
polarize the preference for oom killing either by always preferring a certain
task or completely disabling it. The lowest possible value, -1000, is
equivalent to disabling oom killing entirely for that task since it will always
report a badness score of 0.
--- /etc/init.d/zfs-fuse.orig 2011-01-28 21:20:55.725397214 +0100
+++ /etc/init.d/zfs-fuse 2011-01-28 21:26:32.680999665 +0100
@@ -139,7 +139,7 @@
then
log_end_msg 0
log_action_begin_msg "Immunizing $NAME against OOM kills and sendsigs signals"
- echo -17 > /proc/$(cat /var/run/$NAME.pid)/oom_adj
+ echo -1000 > /proc/$(cat /var/run/$NAME.pid)/oom_score_adj
ES_TO_REPORT=$?
if [ "$ES_TO_REPORT" = 0 ] ; then
log_action_end_msg 0
According to the docs:
The value of /proc/<pid>/oom_score_adj is added to the badness score before it
is used to determine which task to kill. Acceptable values range from -1000
(OOM_SCORE_ADJ_MIN) to +1000 (OOM_SCORE_ADJ_MAX). This allows userspace to
polarize the preference for oom killing either by always preferring a certain
task or completely disabling it. The lowest possible value, -1000, is
equivalent to disabling oom killing entirely for that task since it will always
report a badness score of 0.
Added by
Seth Heeren
on
Jan 28, 2011 09:29 PM
Thanks for the info
On what platforms is this? Is this kernel version specific or distribution specific? I'm trying to establish what initscripts are affected, and since you don't mention your distribution/version I'm having to guess :)
On what platforms is this? Is this kernel version specific or distribution specific? I'm trying to establish what initscripts are affected, and since you don't mention your distribution/version I'm having to guess :)
Added by
(anonymous)
on
Jan 29, 2011 12:14 PM
This is kernel specific - the OOM killer changed in 2.6.36 (see http://lkml.org/lkml/2010/2/15/322)
/proc/pid/oom_adj is now deprecated so that that it may eventually be
removed. The target date for removal is December 2011.
A warning will be printed to the kernel log if a task attempts to use
this interface. Future warning will be suppressed until the kernel is
rebooted to prevent spamming the kernel log.
I'm on Ubuntu 10.10
/proc/pid/oom_adj is now deprecated so that that it may eventually be
removed. The target date for removal is December 2011.
A warning will be printed to the kernel log if a task attempts to use
this interface. Future warning will be suppressed until the kernel is
rebooted to prevent spamming the kernel log.
I'm on Ubuntu 10.10
Added by
Seth Heeren
on
Jan 29, 2011 04:36 PM
Ok thanks for the pointers. It isn't exactly clear to me, what versions of kernels first started supporting the new variant. Depending on that, we might have to stick to the old interface to avoid dropping users behind.
A git log -Soom_score_adj linus/master told me the following bits: it was introduced in v2.6.36, has a 2-year deprecation period, and at least ubuntu seems to have this in it's LTS kernel line since 2.6.37
$ git describe a63d83f42
v2.6.35-5872-ga63d83f
$ git tag --contains a63d83f42 | sort | head -3
Ubuntu-lts-2.6.37-6.17
Ubuntu-lts-2.6.38-1.27
v2.6.36
Log output:
Date: Mon Aug 9 17:19:47 2010 -0700
/proc/pid/oom_adj is now deprecated so that that it may eventually be
removed. The target date for removal is August 2012
commit a63d83f427fbce97a6cea0db2e64b0eb8435cd10
Date: Mon Aug 9 17:19:46 2010 -0700
oom: badness heuristic rewrite
/proc/pid/oom_adj is changed so that its meaning is rescaled into the
units used by /proc/pid/oom_score_adj, and vice versa. ...
A git log -Soom_score_adj linus/master told me the following bits: it was introduced in v2.6.36, has a 2-year deprecation period, and at least ubuntu seems to have this in it's LTS kernel line since 2.6.37
$ git describe a63d83f42
v2.6.35-5872-ga63d83f
$ git tag --contains a63d83f42 | sort | head -3
Ubuntu-lts-2.6.37-6.17
Ubuntu-lts-2.6.38-1.27
v2.6.36
Log output:
Date: Mon Aug 9 17:19:47 2010 -0700
/proc/pid/oom_adj is now deprecated so that that it may eventually be
removed. The target date for removal is August 2012
commit a63d83f427fbce97a6cea0db2e64b0eb8435cd10
Date: Mon Aug 9 17:19:46 2010 -0700
oom: badness heuristic rewrite
/proc/pid/oom_adj is changed so that its meaning is rescaled into the
units used by /proc/pid/oom_score_adj, and vice versa. ...
Added by
(anonymous)
on
Jan 29, 2011 04:59 PM
Maybe one could simply check if the new proc entry exists and use that, and if not use the old one. That should be safe.
Added by
(anonymous)
on
Jan 31, 2011 07:24 AM
Added by
Tim Radke
on
Feb 01, 2012 04:40 AM
According to https://bugzilla.mindrot.org/show_bug.cgi?id=1838 oom_score_adj should be set to -1000 to get the same behavior as oom_adj set to -17.
Added by
Tim Radke
on
Feb 01, 2012 04:55 AM
if [ -f "/proc/$PID/oom_score_adj" ]
then
echo -1000 > "/proc/$PID/oom_score_adj"
else
echo -17 > "/proc/$PID/oom_adj"
fi
then
echo -1000 > "/proc/$PID/oom_score_adj"
else
echo -17 > "/proc/$PID/oom_adj"
fi
Added by
Seth Heeren
on
Feb 01, 2012 05:57 AM
Issue state:
unconfirmed → resolved
Target release:
None → 0.7.0
Responsible manager:
(UNASSIGNED) → sgheeren
Thanks Tim!
things are not always very complicated but the help is very much appreciated. I had never come round to it; The upstart script has been using `oom never` forever:
https://github.com/[…]/zfs.upstart
I have tested this and updated the relevant initscripts under contrib/ and will notify the list on the linked thread.
See your credits:
http://gitweb.zfs-fuse.net/[…]3a1e66f49990ce3523bc9f0a31e
things are not always very complicated but the help is very much appreciated. I had never come round to it; The upstart script has been using `oom never` forever:
https://github.com/[…]/zfs.upstart
I have tested this and updated the relevant initscripts under contrib/ and will notify the list on the linked thread.
See your credits:
http://gitweb.zfs-fuse.net/[…]3a1e66f49990ce3523bc9f0a31e
