* Re: [gentoo-user] cgroupd really do work!
2010-12-05 14:55 ` Florian Philipp
@ 2010-12-05 15:35 ` Mark Knecht
2010-12-05 16:02 ` Florian Philipp
2010-12-07 5:51 ` [gentoo-user] " Scott Prager
2 siblings, 0 replies; 16+ messages in thread
From: Mark Knecht @ 2010-12-05 15:35 UTC (permalink / raw
To: gentoo-user
On Sun, Dec 5, 2010 at 6:55 AM, Florian Philipp
<lists@f_philipp.fastmail.net> wrote:
> Am 04.12.2010 22:00, schrieb Mark Knecht:
>> On Fri, Dec 3, 2010 at 6:16 PM, William Kenworthy <billk@iinet.net.au> wrote:
>>> On Fri, 2010-12-03 at 17:20 -0800, Mark Knecht wrote:
>>>> On Fri, Dec 3, 2010 at 4:41 PM, William Kenworthy <billk@iinet.net.au> wrote:
>>>>> Re the discussion on cgroups on the 24/25 november - on my old AMD
>>>>> barton 2500+ desktop - in the past a load of anything more than 5 made
>>>>> it very painful to use. Add in running windows in a qemu vm at the same
>>>>> time (all in 1.5G ram) and its almost unusable as an interactive
>>>>> desktop.
>>>>>
>>>>> Now, with Florians (thanks for the scripts) version of using it on
>>>>> gentoo, even at a load in excess of 12, its almost as good as with no
>>>>> load for desktop operations!
>>>>>
>>>>> To get the high load I am compiling a new kernel, running win2k in qemu,
>>>>> emerging latest updates, browsing the web and reading email in
>>>>> evolution, running updatedb and other cron jobs (just restarted after a
>>>>> tuxonice hibernate so everything is triggering at once!).
>>>>>
>>>>> Really, really, really neat!
>>>>> BillK
>>>>
> [...]
>>
>> If you're up for it maybe try putting doing a Gentoo Wiki page.
>>
>> Sounds like it's working pretty well. I'd like to try it but I don't
>> have much time to focus on digging out the info.
>>
>> Congrats!
>> -Mark
>>
>
> As you wish:
> http://en.gentoo-wiki.com/wiki/Improve_responsiveness_with_cgroups
>
> I welcome any improvements. This is my first wiki article and English is
> not my mother tongue.
>
> Hope this helps,
> Florian Philipp
>
>
VERY nice start! Thanks!
Cheers,
Mark
^ permalink raw reply [flat|nested] 16+ messages in thread
* Re: [gentoo-user] cgroupd really do work!
2010-12-05 14:55 ` Florian Philipp
2010-12-05 15:35 ` Mark Knecht
@ 2010-12-05 16:02 ` Florian Philipp
2010-12-05 23:54 ` Bill Longman
2010-12-06 20:04 ` [gentoo-user] " Doug Hunley
2010-12-07 5:51 ` [gentoo-user] " Scott Prager
2 siblings, 2 replies; 16+ messages in thread
From: Florian Philipp @ 2010-12-05 16:02 UTC (permalink / raw
To: gentoo-user
[-- Attachment #1: Type: text/plain, Size: 2460 bytes --]
Am 05.12.2010 15:55, schrieb Florian Philipp:
> Am 04.12.2010 22:00, schrieb Mark Knecht:
>> On Fri, Dec 3, 2010 at 6:16 PM, William Kenworthy <billk@iinet.net.au> wrote:
>>> On Fri, 2010-12-03 at 17:20 -0800, Mark Knecht wrote:
>>>> On Fri, Dec 3, 2010 at 4:41 PM, William Kenworthy <billk@iinet.net.au> wrote:
>>>>> Re the discussion on cgroups on the 24/25 november - on my old AMD
>>>>> barton 2500+ desktop - in the past a load of anything more than 5 made
>>>>> it very painful to use. Add in running windows in a qemu vm at the same
>>>>> time (all in 1.5G ram) and its almost unusable as an interactive
>>>>> desktop.
>>>>>
>>>>> Now, with Florians (thanks for the scripts) version of using it on
>>>>> gentoo, even at a load in excess of 12, its almost as good as with no
>>>>> load for desktop operations!
>>>>>
>>>>> To get the high load I am compiling a new kernel, running win2k in qemu,
>>>>> emerging latest updates, browsing the web and reading email in
>>>>> evolution, running updatedb and other cron jobs (just restarted after a
>>>>> tuxonice hibernate so everything is triggering at once!).
>>>>>
>>>>> Really, really, really neat!
>>
>> If you're up for it maybe try putting doing a Gentoo Wiki page.
>>
>> Sounds like it's working pretty well. I'd like to try it but I don't
>> have much time to focus on digging out the info.
>
> As you wish:
> http://en.gentoo-wiki.com/wiki/Improve_responsiveness_with_cgroups
>
> I welcome any improvements. This is my first wiki article and English is
> not my mother tongue.
>
Hmm, I just noticed that something is not right here: Everything works
fine as long as I limit the cgroups to cpu scheduling (`mount -t cgroup
cgroup /dev/cgroup -o cpu`). As soon as I add the blkio subsystem for
disk I/O scheduling ("-o cpu,blkio" or no "-o" at all), I can no longer
create cgroup hierarchies. Only the top level "user" cgroup is accepted.
Another issue is that the kernel documentation in
/usr/src/linux/Documentation/cgroups/cgroups.txt specifies that the
release_agent can also be included as a mount option with
`mount -t cgroup cgroup /dev/cgroup -o
cpu,release_agent='/usr/local/sbin/cgroup_clean'`
That seems to be no valid option, though. At least on my system it
causes mount to fail. Unfortunately there is no output on dmesg.
Can someone else reproduce this? I'm on gentoo-sources-2.6.34-r12.
Thanks in advance!
Florian Philipp
[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 262 bytes --]
^ permalink raw reply [flat|nested] 16+ messages in thread* Re: [gentoo-user] cgroupd really do work!
2010-12-05 16:02 ` Florian Philipp
@ 2010-12-05 23:54 ` Bill Longman
2010-12-06 4:51 ` Florian Philipp
2010-12-06 20:04 ` [gentoo-user] " Doug Hunley
1 sibling, 1 reply; 16+ messages in thread
From: Bill Longman @ 2010-12-05 23:54 UTC (permalink / raw
To: gentoo-user
[-- Attachment #1: Type: text/plain, Size: 1072 bytes --]
>
> Hmm, I just noticed that something is not right here: Everything works
> fine as long as I limit the cgroups to cpu scheduling (`mount -t cgroup
> cgroup /dev/cgroup -o cpu`). As soon as I add the blkio subsystem for
> disk I/O scheduling ("-o cpu,blkio" or no "-o" at all), I can no longer
> create cgroup hierarchies. Only the top level "user" cgroup is accepted.
>
> Another issue is that the kernel documentation in
> /usr/src/linux/Documentation/cgroups/cgroups.txt specifies that the
> release_agent can also be included as a mount option with
> `mount -t cgroup cgroup /dev/cgroup -o
> cpu,release_agent='/usr/local/sbin/cgroup_clean'`
> That seems to be no valid option, though. At least on my system it
> causes mount to fail. Unfortunately there is no output on dmesg.
>
> Can someone else reproduce this? I'm on gentoo-sources-2.6.34-r12.
>
>
I had this very same problem when I was trying this last year, Florian. I
didn't pursue it any further so I can't tell you any solutions. I do
remember it though because it was quite frustrating.
--
Bill Longman
[-- Attachment #2: Type: text/html, Size: 1417 bytes --]
^ permalink raw reply [flat|nested] 16+ messages in thread* Re: [gentoo-user] cgroupd really do work!
2010-12-05 23:54 ` Bill Longman
@ 2010-12-06 4:51 ` Florian Philipp
0 siblings, 0 replies; 16+ messages in thread
From: Florian Philipp @ 2010-12-06 4:51 UTC (permalink / raw
To: gentoo-user
[-- Attachment #1: Type: text/plain, Size: 1455 bytes --]
Am 06.12.2010 00:54, schrieb Bill Longman:
>
>
> Hmm, I just noticed that something is not right here: Everything works
> fine as long as I limit the cgroups to cpu scheduling (`mount -t cgroup
> cgroup /dev/cgroup -o cpu`). As soon as I add the blkio subsystem for
> disk I/O scheduling ("-o cpu,blkio" or no "-o" at all), I can no longer
> create cgroup hierarchies. Only the top level "user" cgroup is accepted.
>
> Another issue is that the kernel documentation in
> /usr/src/linux/Documentation/cgroups/cgroups.txt specifies that the
> release_agent can also be included as a mount option with
> `mount -t cgroup cgroup /dev/cgroup -o
> cpu,release_agent='/usr/local/sbin/cgroup_clean'`
> That seems to be no valid option, though. At least on my system it
> causes mount to fail. Unfortunately there is no output on dmesg.
>
> Can someone else reproduce this? I'm on gentoo-sources-2.6.34-r12.
>
>
> I had this very same problem when I was trying this last year, Florian.
> I didn't pursue it any further so I can't tell you any solutions. I do
> remember it though because it was quite frustrating.
>
> --
> Bill Longman
Thanks for the answer!
I think I've found the answer, at least for the first issue: Look at
this bug: https://bugzilla.kernel.org/show_bug.cgi?id=15980
I still don't know what causes the second issue.
Regards,
Florian Philipp
[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 262 bytes --]
^ permalink raw reply [flat|nested] 16+ messages in thread
* [gentoo-user] Re: cgroupd really do work!
2010-12-05 16:02 ` Florian Philipp
2010-12-05 23:54 ` Bill Longman
@ 2010-12-06 20:04 ` Doug Hunley
2010-12-06 20:44 ` Doug Hunley
2010-12-06 22:00 ` Florian Philipp
1 sibling, 2 replies; 16+ messages in thread
From: Doug Hunley @ 2010-12-06 20:04 UTC (permalink / raw
To: gentoo-user
[-- Attachment #1: Type: text/plain, Size: 944 bytes --]
Thanks a ton for the wiki page! I 'smartened' up the scripts a bit to deal
with the various locations and kernel versions:
cgroup_clean:
#!/bin/sh
if [ -d /sys/fs/cgroup ] ; then
cdir=/sys/fs/cgroup
else
cdir=/dev/cgroup
fi
rmdir /dev/cgroup/$*
cgroup_start:
#!/bin/sh
if [ -d /sys/fs/cgroup ] ; then
cdir=/sys/fs/cgroup
else
cdir=/dev/cgroup
mkdir $cdir
fi
kern_version=`uname -r|cut -d\- -f1|tr -d '.'`
if [ $kern_version -lt 2638 ] ; then
mount -t cgroup cgroup $cdir -o cpu
else
mount -t cgroup cgroup $cdir -o cpu,blkio
fi
mkdir -m 0777 $cdir/user
/bin/echo '/root/bin/cgroup_clean' > $cdir/release_agent
bashrc:
# Turn on cgroups
if [ "$PS1" ] ; then
if [ -d /sys/fs/cgroup ] ; then
cdir=/sys/fs/cgroup
else
cdir=/dev/cgroup
fi
mkdir -p -m 0700 $cdir/user/$$ > /dev/null 2>&1
/bin/echo $$ > $cdir/user/$$/tasks
/bin/echo '1' > $cdir/user/$$/notify_on_release
fi
Hope someone finds that useful.
[-- Attachment #2: Type: text/html, Size: 1453 bytes --]
^ permalink raw reply [flat|nested] 16+ messages in thread* [gentoo-user] Re: cgroupd really do work!
2010-12-06 20:04 ` [gentoo-user] " Doug Hunley
@ 2010-12-06 20:44 ` Doug Hunley
2010-12-06 22:00 ` Florian Philipp
1 sibling, 0 replies; 16+ messages in thread
From: Doug Hunley @ 2010-12-06 20:44 UTC (permalink / raw
To: gentoo-user
[-- Attachment #1: Type: text/plain, Size: 253 bytes --]
On Mon, Dec 6, 2010 at 15:04, Doug Hunley <doug@hunley.homeip.net> wrote:
> cgroup_clean:
> #!/bin/sh
> if [ -d /sys/fs/cgroup ] ; then
> cdir=/sys/fs/cgroup
> else
> cdir=/dev/cgroup
> fi
>
> rmdir /dev/cgroup/$*
>
change that to: rmdir $cdir/$*
[-- Attachment #2: Type: text/html, Size: 918 bytes --]
^ permalink raw reply [flat|nested] 16+ messages in thread
* Re: [gentoo-user] Re: cgroupd really do work!
2010-12-06 20:04 ` [gentoo-user] " Doug Hunley
2010-12-06 20:44 ` Doug Hunley
@ 2010-12-06 22:00 ` Florian Philipp
1 sibling, 0 replies; 16+ messages in thread
From: Florian Philipp @ 2010-12-06 22:00 UTC (permalink / raw
To: gentoo-user
[-- Attachment #1: Type: text/plain, Size: 562 bytes --]
Am 06.12.2010 21:04, schrieb Doug Hunley:
> Thanks a ton for the wiki page! I 'smartened' up the scripts a bit to
> deal with the various locations and kernel versions:
[...]
>
> Hope someone finds that useful.
Sure, nice extension. Since it is your code, please apply it to the wiki
page yourself. You don't need my approval for this. ;)
One slight improvement: Add a line `unset -v cdir` to the end of the
.bashrc inclusion. cdir is not needed afterwards but will remain as a
shell variable if it is not deleted.
Regards,
Florian Philipp
[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 262 bytes --]
^ permalink raw reply [flat|nested] 16+ messages in thread
* Re: [gentoo-user] cgroupd really do work!
2010-12-05 14:55 ` Florian Philipp
2010-12-05 15:35 ` Mark Knecht
2010-12-05 16:02 ` Florian Philipp
@ 2010-12-07 5:51 ` Scott Prager
2010-12-07 8:43 ` Neil Bothwick
2 siblings, 1 reply; 16+ messages in thread
From: Scott Prager @ 2010-12-07 5:51 UTC (permalink / raw
To: gentoo-user
[-- Attachment #1: Type: text/plain, Size: 848 bytes --]
On Sun, Dec 5, 2010 at 9:55 AM, Florian Philipp <
lists@f_philipp.fastmail.net> wrote:
http://en.gentoo-wiki.com/wiki/Improve_responsiveness_with_cgroups
>
> I welcome any improvements. This is my first wiki article and English is
> not my mother tongue.
>
> Hope this helps,
> Florian Philipp
>
>
This is my first post on this mailing list, so please correct me if this is
in any way the wrong place.
I found that /etc/conf.d/local.start was not on my system like the article
seems to expect. /etc/conf.d/local local was and it already had two
functions, local_start and _stop, which seemed to be what local.start was
supposed to be. And using /etc/conf.d/local.start did not work
whereas /etc/conf.d/local did.
Did i do something wrong in fallowing the article, do some systems use
local.start and some just local, or is there something else?
[-- Attachment #2: Type: text/html, Size: 1406 bytes --]
^ permalink raw reply [flat|nested] 16+ messages in thread
* Re: [gentoo-user] cgroupd really do work!
2010-12-07 5:51 ` [gentoo-user] " Scott Prager
@ 2010-12-07 8:43 ` Neil Bothwick
0 siblings, 0 replies; 16+ messages in thread
From: Neil Bothwick @ 2010-12-07 8:43 UTC (permalink / raw
To: gentoo-user
[-- Attachment #1: Type: text/plain, Size: 707 bytes --]
On Tue, 7 Dec 2010 00:51:10 -0500, Scott Prager wrote:
> I found that /etc/conf.d/local.start was not on my system like the
> article seems to expect. /etc/conf.d/local local was and it already had
> two functions, local_start and _stop, which seemed to be what
> local.start was supposed to be. And using /etc/conf.d/local.start did
> not work whereas /etc/conf.d/local did.
Baselayout1 uses local.start and local.stop, baselayout 2 uses start and
stop functions in a single local file. AFAIR baselayout 2 will look for
local.start and local.stop, but only if local does not exist, or it used
to.
--
Neil Bothwick
Teamwork is essential; it gives the enemy other people to shoot at.
[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 198 bytes --]
^ permalink raw reply [flat|nested] 16+ messages in thread