[Netkit.users] First official release of Netkit-NG

Saverio Proto zioproto at gmail.com
Fri Jun 13 14:54:27 CEST 2014


Hello,

sorry, noatime was not really solving the problem :( It was a false positive.

The problem is the access from the host machine to the netkit
filesystem, when the UML machine is doing disk intesive operations.
There are some operations that cause the journaling of the host file
system to get all the IO resources.

To check the worst case just do

vstart nuovo -M 512 --eth0=tap,172.17.0.1,172.17.0.2 -W

set a nameserver in /etc/resolv.conf and try "apt-get update"
in my case the UML system hangs for more than 4 minutes on "Reading
package lists".

To make sure it is filesystem related I did this test, I moved the
netkit-fs to a ramdisk.
Doing the same:
vstart nuovo -M 512 --eth0=tap,172.17.0.1,172.17.0.2 -W
apt-get update (in the UML machine)
this is now SUPER fast.

I append here my bash history.

Saverio


 2088  sudo mount -t tmpfs -o size=2512M tmpfs /mnt/ramdisk
 2089  cd
 2090  rm netkit-ng/fs/netkit-fs-i386-F7.0
 2091  tar -jxvf ~/Scaricati/netkit-ng-filesystem-i386-F7.0.tar.bz2
 2092  cp netkit-ng/fs/netkit-fs-i386-F7.0 /mnt/ramdisk/
 2093  cd netkit-ng/fs/
 2094  ls -alh
 2095  rm netkit-fs
 2096  ln -s /mnt/ramdisk/netkit-fs-i386-F7.0 netkit-fs
 2097  ls -alh
 2098  cd
 2099  cd /tmp/
 2100  vstart nuovo -M 512 --eth0=tap,172.17.0.1,172.17.0.2 -W

2014-06-13 14:18 GMT+02:00 Julien Iguchi-Cartigny
<julien.iguchi-cartigny a lifl.fr>:
> On 06/13/2014 02:16 PM, Saverio Proto wrote:
>> I found the problem.
>>
>> http://tldp.org/LDP/solrhe/Securing-Optimizing-Linux-RH-Edition-v1.3/chap6sec73.html
>>
>> my home partition is NOT mounted with noatime
>>
>> I just tried
>>
>> chattr +A netkit-fs-i386-F7.0
>>
>> saverio a nockid:~/netkit-ng/fs$ lsattr netkit-fs-i386-F7.0
>> -------A-----e- netkit-fs-i386-F7.0
>> saverio a nockid:~/netkit-ng/fs$
>>
>>
>> I started a machine with -W (Cow mode) and it is very fast now.
>>
>> I will make more detailed tests later, but this looks like to be the problem.
>
> Ok I'm waiting your feedback then I will write a new test in check
> configuration to avoid it again.
>
> Thanks again for your precious help.
>
> Julien
>
>> Saverio
>>
>>
>>
>> 2014-06-13 14:03 GMT+02:00 Saverio Proto <zioproto a gmail.com>:
>>> I mean jbd2 is taking IO time on the host machine.
>>>
>>> Saverio
>>>
>>>
>>> 2014-06-13 13:10 GMT+02:00 Julien Iguchi-Cartigny
>>> <julien.iguchi-cartigny a lifl.fr>:
>>>> On 06/13/2014 01:07 PM, Saverio Proto wrote:
>>>>> Using iotop I see that the process that consumes all the IO time when
>>>>> the UML machines are very slow is:
>>>>>
>>>>> jbd2/sdb1-8
>>>>
>>>> jbd2 is now used (it is a fs system). I will try to build new images
>>>> this WE without it.
>>>>
>>>> Thanks for your feedback.
>>>>
>>>> J.
>>>>
>>>>
>>>>> I have this process using 99.99% IO time
>>>>>
>>>>> This is when the UML kernel needs to perform a disk write operation
>>>>>
>>>>> Saverio
>>>>>
>>>>>
>>>>> 2014-06-13 13:01 GMT+02:00 Saverio Proto <zioproto a gmail.com>:
>>>>>>> Could you precise if it's the I/O or the CPU use which seems slow ?
>>>>>>
>>>>>> Hello,
>>>>>> it is I/O problem. CPU is running low.
>>>>>>
>>>>>>> Could you describe your test computer architecture  (32/64 bits ?
>>>>>>> distribution ?) ?
>>>>>>
>>>>>> saverio a nockid:~$ cat /etc/issue
>>>>>> Ubuntu 12.04.4 LTS \n \l
>>>>>>
>>>>>> saverio a nockid:~$ uname -a
>>>>>> Linux nockid 3.2.0-58-generic #88-Ubuntu SMP Tue Dec 3 17:37:58 UTC
>>>>>> 2013 x86_64 x86_64 x86_64 GNU/Linux
>>>>>> saverio a nockid:~$ cat /proc/cpuinfo | head
>>>>>> processor    : 0
>>>>>> vendor_id    : GenuineIntel
>>>>>> cpu family    : 6
>>>>>> model        : 42
>>>>>> model name    : Intel(R) Core(TM) i7-2600 CPU @ 3.40GHz
>>>>>> stepping    : 7
>>>>>> microcode    : 0x17
>>>>>> cpu MHz        : 2400.000
>>>>>> cache size    : 8192 KB
>>>>>> physical id    : 0
>>>>>> saverio a nockid:~$
>>>>> _______________________________________________
>>>>> Netkit.users mailing list
>>>>> Netkit.users a list.dia.uniroma3.it
>>>>> http://list.dia.uniroma3.it/mailman/listinfo/netkit.users
>>>>>
>>>>
>>>> _______________________________________________
>>>> Netkit.users mailing list
>>>> Netkit.users a list.dia.uniroma3.it
>>>> http://list.dia.uniroma3.it/mailman/listinfo/netkit.users
>> _______________________________________________
>> Netkit.users mailing list
>> Netkit.users a list.dia.uniroma3.it
>> http://list.dia.uniroma3.it/mailman/listinfo/netkit.users
>>
>
> _______________________________________________
> Netkit.users mailing list
> Netkit.users a list.dia.uniroma3.it
> http://list.dia.uniroma3.it/mailman/listinfo/netkit.users


More information about the Netkit.users mailing list