Hello Jan,
After doing analysis, I found that When memory hog process (mem_process) tries
to mmap double the memory of the cgroup limit which is 4096 . I am getting OOM
in that situation and send kill signal to process (which is expected behaviour)
. But it could not kill the process and
it sets the process state to uninterrupted sleep state.
To get this OOM , i need to change oom_adj value of the process '208' from -17
to -16. If i dont do this, I dont get OOM and process always stuck at mmap and
try to allocate 8192 Bytes of memory.
logs:-
[2-74.5749] Memory cgroup out of memory: Kill process 208 (mem_process) score
147000 or sacrifice child
[2-74.5840] Killed process 208 (mem_process) total-vm:2704kB, anon-rss:80kB,
file-rss:484kB
by mmap is -1237753856
Warning: bad syntax, perhaps a bogus '-'? See /usr/share/doc/procps-3.2.8/FAQ
root 208 39.6 0.0 2704 564 ? D 00:01 0:02 mem_process -m
8192 // D - means uninterrupted sleep
root 215 0.0 0.0 2412 584 ? S 00:01 0:00 grep -r 208
I am using 3.10.84 kernel. LTP is of latest version (ltp-full-20150420)
Regards
Manjeet
------- Original Message -------
Sender : Jan Stancek<jstan...@redhat.com>
Date : Jul 27, 2015 17:32 (GMT+09:00)
Title : Re: [LTP] Regarding memcg control testcase (mem_process.c)
----- Original Message -----
> From: "Manjeet Pawar"
> To: "Jan Stancek"
> Cc: ltp-list@lists.sourceforge.net, "AJEET YADAV" , "PANKAJ MISHRA"
> Sent: Monday, 27 July, 2015 10:12:46 AM
> Subject: Re: Re: [LTP] Regarding memcg control testcase (mem_process.c)
>
> Hello Mr Jan,
> Thanks for your reply,
> With actual testscript I checked my "$!" , it is printing correct pid.
> Also changes which you suggest, I tried with those changes , it is showing me
> either "mmap failed:ffffffff"
When it fails, what is the errno?
> sometimes or "a successful mmap call".
Where is this message coming from? I can't find such string in my LTP.
> In my case "memory.usage_in_bytes" is 4096.
> Testcase is run as "./memcg_control_test.sh 4096 4096 8192".
> I am not getting this "mem_process getting killed by OOM seems like expected
> outcome:"
> May i know which kernel verison you are using.? Now I am doubtful on my
> kernel.
I'm on 4.0.4.
What is your kernel/arch and LTP version?
Regards,
Jan
>
> Regards
> Manjeet
>
> ------- Original Message -------
> Sender : Jan Stancek
> Date : Jul 26, 2015 20:28 (GMT+09:00)
> Title : Re: [LTP] Regarding memcg control testcase (mem_process.c)
>
>
> ----- Original Message -----
> > From: "Manjeet Pawar"
> > To: ltp-list@lists.sourceforge.net
> > Cc: "ajeet y" , "pankaj m"
> > Sent: Friday, 24 July, 2015 6:07:45 AM
> > Subject: [LTP] Regarding memcg control testcase (mem_process.c)
> >
> > Hi,
> >
> > I am trying to run memcg control testcase using script
> > memcg_control_test.sh.
> > This script calls a memory hog process 'mem_process'.
> > The condition for this testcase to pass is 'when the memory hog process is
> > killed on crossing boundary'. please check the below function.
> >
> >
> > -----------------------------
> > # Check if the test process is killed on crossing boundary
> > test_proc_kill()
> > {
> > ( cd $TMP && mem_process -m $PROC_MEM & )
> > sleep 1
> > echo $! > tasks
>
> mem_process is getting started via subshell, so I'd assume $! in parent is
> bogus.
> Can you try this instead:
>
> pushd $TMP || return
> mem_process -m $PROC_MEM &
> local pid=$!
> popd
>
> sleep 1
> echo $pid > tasks
>
>
> >
> > #Instruct the test process to start acquiring memory
> > echo m > $STATUS_PIPE
> > sleep 5
> >
> > #Check if killed
> > ps -p $! > /dev/null 2> /dev/null ----> at this point the process
> > 'mem_process' must have terminated so that last return value ($?) can
> > be
> > non-zero.
> > if [ $? -eq 0 ]; then
> > echo m > $STATUS_PIPE
> > echo x > $STATUS_PIPE
> > else
> > : $((KILLED_CNT += 1))
> > fi
> > }
> > -------------------------
> > In my case: 'mem_process' does not terminate until we pass 'x' to it, but
> > the
> > expected behaviour according to testcase that the memory hog process
> > must be killed on crossing bounday.
> > Inputs taken by testcase are : "./memcg_control_test.sh 4096 4096 8192"
> > If I change the inputs from 8192 to other (1024,10MB,100MB etc) . In all
> > situation memory hog process does not terminate.
> > So what is the ideal condition for this process to terminate to make whole
> > testcase pass. ?
>
> mem_process getting killed by OOM seems like expected outcome:
>
> memcg_control 0 TINFO : Test #1: Checking if the memory usage limit
> imposed by the topmost group is enforced
> ./memcg_control_test.sh: line 56: 2795 Killed mem_process
> -m $PROC_MEM (wd: /tmp)
> memcg_control 1 TPASS : Test #1: passed
> memcg_control 1 TPASS : memcg_control: passed
>
> If it still doesn't get killed, I'd recommend checking
> "memory.usage_in_bytes".
>
> Regards,
> Jan
>
>
> >
> > ------------------------------------------------------------------------------
> > _______________________________________________
> > Ltp-list mailing list
> > Ltp-list@lists.sourceforge.net
> > https://lists.sourceforge.net/lists/listinfo/ltp-list
> >
------------------------------------------------------------------------------
_______________________________________________
Ltp-list mailing list
Ltp-list@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/ltp-list
