oom killer and its superior braindamage in 2.4

oom killer and its superior braindamage in 2.4

[David Mansfield]

Marc, Rik,

> - Feb 21 10:04:57 codeman kernel: Out of Memory: Killed process 2657 
> (apache).
>
> The above log entry (apache) appeared for about 4 hours every some 
> seconds (same PID) until I thought about sysrq-b to get out of this 
> braindead behaviour. The machine was somewhat dead for me because I was 
> not able to do anything but sysrq. The system itself was _not_ dead, 
> there was massive disk i/o. This is 2.4.20 vanilla.

This exact thing happened to me as well, on a 2.4.20-pre that hasn't been 
upgraded to 2.4.20 yet.  The thing that concerns me most is:

Why won't the system kill the process it claims to be killing?

If, in Marc's case, the system wants to kill PID 2657, a lowly sleeping 
apache process, why can't it?  This is a bug for sure.

For me, there was some python process chosen as the one for killing and it 
repeated the 'Out of Memory: Killed process xxxxx (python)' for hours 
while making no progress.  The machine was still routing packets but I 
couldn't log in.  Sys-rq was disabled, so I was forced to use the big red 
button.

Rik, any ideas?

David

-- 
/==============================\
| David Mansfield              |
| lkml@dm.cobite.com           |
\==============================/

Re: oom killer and its superior braindamage in 2.4

[Faik Uygur]

> This exact thing happened to me as well, on a 2.4.20-pre that hasn't been 
> upgraded to 2.4.20 yet.  The thing that concerns me most is:
> 
> Why won't the system kill the process it claims to be killing?
> 
> If, in Marc's case, the system wants to kill PID 2657, a lowly sleeping 
> apache process, why can't it?  This is a bug for sure.
> 
> For me, there was some python process chosen as the one for killing and it 
> repeated the 'Out of Memory: Killed process xxxxx (python)' for hours 
> while making no progress.  The machine was still routing packets but I 
> couldn't log in.  Sys-rq was disabled, so I was forced to use the big red 
> button.
> 
> Rik, any ideas?

But, did you follow that thread? Rik van Riel, already suggested a solution for
the problem.

http://marc.theaimsgroup.com/?l=linux-kernel&m=104594301523518&w=2

Re: oom killer and its superior braindamage in 2.4

[Rik van Riel]

On Sun, 23 Feb 2003, David Mansfield wrote:

> Rik, any ideas?

You could try the patch I sent to Marc and linux-kernel
yesterday afternoon ;)

Rik
-- 
Engineers don't grow up, they grow sideways.
http://www.surriel.com/		http://kernelnewbies.org/

Re: oom killer and its superior braindamage in 2.4

[David Mansfield]

> On Sun, 23 Feb 2003, David Mansfield wrote:
> 
> > Rik, any ideas?
> 
> You could try the patch I sent to Marc and linux-kernel
> yesterday afternoon ;)
> 

You miss my point completely.  The kernel has ALREADY chosen a task to 
kill.  I don't care to adjust the 'badness' function.  The kernel has 
already chosen a bad task.

If you read my post, the bug is that the kernel CANNOT kill that process?  
Why?  If it's really a bad process, shouldn't it be the one that gets 
killed? 

With you patch we have:

1) Kernel goes OOM
2) Kernel picks the worst task to kill using badness()
3) Kernel attempts to kill this task but fails due to some {reason|bug}.
4) Kernel now picks some other task to kill even though the 'baddest' one 
is allowed to hang out.

This is my question, and I don't see how the patch addresses it.

David



-- 
/==============================\
| David Mansfield              |
| lkml@dm.cobite.com           |
\==============================/

Re: oom killer and its superior braindamage in 2.4

[Rik van Riel]

On Sun, 23 Feb 2003, David Mansfield wrote:

> If you read my post, the bug is that the kernel CANNOT kill that
> process?  Why?  If it's really a bad process, shouldn't it be the one
> that gets killed?

> This is my question, and I don't see how the patch addresses it.

And you won't see one, either.  You cannot change the
semantics of uninterruptible sleep, nor can the OOM
killer change other device driver things.

This means the OOM killer has little choice but to
"hope for the best" and pick another process if the
first process chosen can't exit.

If you think you can fix all drivers to work fine
when tasks suddenly disappear, I guess you might
wnat to create such a patch ...

regards,

Rik
-- 
Engineers don't grow up, they grow sideways.
http://www.surriel.com/		http://kernelnewbies.org/

Re: oom killer and its superior braindamage in 2.4

[David Mansfield]

> 
> > If you read my post, the bug is that the kernel CANNOT kill that
> > process?  Why?  If it's really a bad process, shouldn't it be the one
> > that gets killed?
> 
> > This is my question, and I don't see how the patch addresses it.
> 
> And you won't see one, either.  You cannot change the
> semantics of uninterruptible sleep, nor can the OOM
> killer change other device driver things.

So you're saying that a process can stay in the D state, without ever 
getting enough resources to complete a single Uninteruptible wait, for 
hours at a time?

Ok.  Now I understand your patch.  Thanks for the info.

You should push your patch to Marcelo. 

Thanks,
David

-- 
/==============================\
| David Mansfield              |
| david@cobite.com             |
\==============================/

Re: oom killer and its superior braindamage in 2.4

[Rik van Riel]

On Sun, 23 Feb 2003, David Mansfield wrote:

> So you're saying that a process can stay in the D state, without ever
> getting enough resources to complete a single Uninteruptible wait, for
> hours at a time?

Or even in the R state, but that would only happen when there
is a kernel bug.  The OOM killer can do nothing but hope for
the best and try another process if the first one doesn't want
to exit.

> Ok.  Now I understand your patch.  Thanks for the info.
>
> You should push your patch to Marcelo.

Will do.

cheers,

Rik
-- 
Engineers don't grow up, they grow sideways.
http://www.surriel.com/		http://kernelnewbies.org/

Re: oom killer and its superior braindamage in 2.4

[Marc-Christian Petersen]

On Monday 24 February 2003 10:13, Mikael Starvik wrote:

Hi Mikael,

> Does everyone agree that killing a process is always the best approach
> to resolve an OOM? If the OOM is caused by e.g. a growing tmpfs or
> memory leaks in the kernel it won't help much to kill processes that
> may respawn.
Well, I don't agree that it's always the best approach. Other bad things, you 
metioned it, can happen.

> Would it be useful if it was possible to register another oom-handler?
> Some architectures could then choose to e.g. reboot the system instead.
I'd like to see _an option_ (read: not default but an option, e.g. boot 
parameter) that will reboot the machine after $specified_time if an OOM 
killing action does not stop.

ciao, Marc

RE: oom killer and its superior braindamage in 2.4

[Mikael Starvik]

Does everyone agree that killing a process is always the best approach
to resolve an OOM? If the OOM is caused by e.g. a growing tmpfs or 
memory leaks in the kernel it won't help much to kill processes that
may respawn. 

Would it be useful if it was possible to register another oom-handler?
Some architectures could then choose to e.g. reboot the system instead.

/Mikael

-----Original Message-----
From: linux-kernel-owner@vger.kernel.org
[mailto:linux-kernel-owner@vger.kernel.org]On Behalf Of Marc-Christian
Petersen
Sent: Saturday, February 22, 2003 8:35 PM
To: linux-kernel@vger.kernel.org
Subject: oom killer and its superior braindamage in 2.4


Hi all,

I just thought (ok it was yesterday) about stress testing my mysql db.
I used this:
- mystress.pl localhost mysql root test 600 300 60 "select * from user"

It worked like a charme. So I tried:
- mystress.pl localhost mysql root test 1800 900 60 "select * from user"

My machine has 512MB RAM and 512MB SWAP.

I expected that the 2nd run will OOM my machine but I did not expect this 
silly behaviour.

The following log entry appeared only _once_ (there were ~700 mysqld running)

- Feb 21 10:03:22 codeman kernel: Out of Memory: Killed process 1463 (mysqld).


Instead of really killing either mysqld or mystress.pl the OOM killer decided 
to kill apache (apache did nothing but had 5 threads sleeping)

- Feb 21 10:04:57 codeman kernel: Out of Memory: Killed process 2657 (apache).

The above log entry (apache) appeared for about 4 hours every some seconds 
(same PID) until I thought about sysrq-b to get out of this braindead 
behaviour. The machine was somewhat dead for me because I was not able to do 
anything but sysrq. The system itself was _not_ dead, there was massive disk 
i/o. This is 2.4.20 vanilla.

Is there any chance we can fix this up?

ciao, Marc


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Re: oom killer and its superior braindamage in 2.4

[Marc-Christian Petersen]

On Sunday 23 February 2003 21:18, Rik van Riel wrote:

Hi Rik,

> It'd be interesting to know where these processes are spending
> their CPU time and why they're not catching their signals.
I'll look into it again when I do the next run.

> > Sysrq-i gave me the chance to get out of the OOM killing process and
> > only kernel threads were left + getty's so I was able to log in again.
> Strange, so sysrq-i manages to kill the processes, but the OOM
> killer doesn't kill the processes ?
yep, so it is.

> This is very suspect because the OOM killer uses force_sig in
> the same way the sysrq-i handler does...
indeed. Well, sysrq-i need about 5 seconds to give me my getty back.

Anyway, your patch should go into -BK. Your patch does _not_ introduce this 
behaviour, it's present even w/o your patch but your approach makes things 
better :)

ciao, Marc

Re: oom killer and its superior braindamage in 2.4

[Rik van Riel]

On Sun, 23 Feb 2003, Marc-Christian Petersen wrote:

> > Does the below patch fix your problem ?

> With your patch, mystress.pl was marked to get killed, every PID only
> once, no apache or similar (good). ... But the strange thing is, that it
> seems none of the processes, which are marked to be killed, get killed.
> So sysrq-t tells me.

It'd be interesting to know where these processes are spending
their CPU time and why they're not catching their signals.

> Sysrq-i gave me the chance to get out of the OOM killing process and
> only kernel threads were left + getty's so I was able to log in again.

Strange, so sysrq-i manages to kill the processes, but the OOM
killer doesn't kill the processes ?

This is very suspect because the OOM killer uses force_sig in
the same way the sysrq-i handler does...

regards,

Rik
-- 
Engineers don't grow up, they grow sideways.
http://www.surriel.com/		http://kernelnewbies.org/

Re: oom killer and its superior braindamage in 2.4

[Marc-Christian Petersen]

On Saturday 22 February 2003 21:32, Rik van Riel wrote:

Hi Rik,

> > > - Feb 21 10:04:57 codeman kernel: Out of Memory: Killed process 2657
> > > (apache).
> > > The above log entry (apache) appeared for about 4 hours every some
> > > seconds (same PID) until I thought about sysrq-b
> > > Is there any chance we can fix this up?
> > Yes.
> Never mind my last idea, it can be done much simpler ;)
hehe :)

> Does the below patch fix your problem ?
Well, this makes a difference. I filled up my memory with something else 
before starting mystress.pl because of top's|ps' slowness with many processes.
I had about 400 processes. The test from yesterday had ~ 1800.

With your patch, mystress.pl was marked to get killed, every PID only once, no 
apache or similar (good). ... But the strange thing is, that it seems none of 
the processes, which are marked to be killed, get killed. So sysrq-t tells 
me. Sysrq-i gave me the chance to get out of the OOM killing process and only 
kernel threads were left + getty's so I was able to log in again.

ciao, Marc

Re: oom killer and its superior braindamage in 2.4

[Rik van Riel]

On Sat, 22 Feb 2003, Rik van Riel wrote:
> On Sat, 22 Feb 2003, Marc-Christian Petersen wrote:
>
> > - Feb 21 10:04:57 codeman kernel: Out of Memory: Killed process 2657 (apache).
> >
> > The above log entry (apache) appeared for about 4 hours every some
> > seconds (same PID) until I thought about sysrq-b
>
> > Is there any chance we can fix this up?
>
> Yes.

Never mind my last idea, it can be done much simpler ;)

Does the below patch fix your problem ?

Rik
-- 
Engineers don't grow up, they grow sideways.
http://www.surriel.com/		http://kernelnewbies.org/


===== mm/oom_kill.c 1.11 vs edited =====
--- 1.11/mm/oom_kill.c	Fri Aug 16 10:59:46 2002
+++ edited/mm/oom_kill.c	Sat Feb 22 17:31:49 2003
@@ -61,6 +61,9 @@

 	if (!p->mm)
 		return 0;
+
+	if (p->flags & PF_MEMDIE)
+		return 0;
 	/*
 	 * The memory size of the process is the basis for the badness.
 	 */

Re: oom killer and its superior braindamage in 2.4

[Rik van Riel]

On Sat, 22 Feb 2003, Marc-Christian Petersen wrote:

> - Feb 21 10:04:57 codeman kernel: Out of Memory: Killed process 2657 (apache).
>
> The above log entry (apache) appeared for about 4 hours every some
> seconds (same PID) until I thought about sysrq-b

> Is there any chance we can fix this up?

Yes.

1) add a VM_KILLED flag
2) set this flag on the p->mm->def_flags when you kill a
   process/thread from oom_kill.c
3) clear the flag on process exit

4) on a new call to oom_kill, skip processes when
   (p->mm->def_flags & VM_KILLED) by returning 0
   points for such a process

cheers,

Rik
-- 
Engineers don't grow up, they grow sideways.
http://www.surriel.com/		http://kernelnewbies.org/

oom killer and its superior braindamage in 2.4

[Marc-Christian Petersen]

Hi all,

I just thought (ok it was yesterday) about stress testing my mysql db.
I used this:
- mystress.pl localhost mysql root test 600 300 60 "select * from user"

It worked like a charme. So I tried:
- mystress.pl localhost mysql root test 1800 900 60 "select * from user"

My machine has 512MB RAM and 512MB SWAP.

I expected that the 2nd run will OOM my machine but I did not expect this 
silly behaviour.

The following log entry appeared only _once_ (there were ~700 mysqld running)

- Feb 21 10:03:22 codeman kernel: Out of Memory: Killed process 1463 (mysqld).


Instead of really killing either mysqld or mystress.pl the OOM killer decided 
to kill apache (apache did nothing but had 5 threads sleeping)

- Feb 21 10:04:57 codeman kernel: Out of Memory: Killed process 2657 (apache).

The above log entry (apache) appeared for about 4 hours every some seconds 
(same PID) until I thought about sysrq-b to get out of this braindead 
behaviour. The machine was somewhat dead for me because I was not able to do 
anything but sysrq. The system itself was _not_ dead, there was massive disk 
i/o. This is 2.4.20 vanilla.

Is there any chance we can fix this up?

ciao, Marc