Problems with XFS in a power failure

Problems with XFS in a power failure

[Uwe Kleine-König]

Hello,

my machine suffered a power failure while doing a apt-get upgrade.

This damaged several files.  E.g.

	root@cepheus:~# xxd /var/lib/dpkg/info/myspell-en-us.postrm
	0000000: 0000 0000 0000 0000 0000 0000 0000 0000  ................
	0000010: 0000 0000 0000 0000 0000 0000 0000 0000  ................
	0000020: 0000 0000 0000 0000 0000 0000 0000 0000  ................
	0000030: 0000 0000 0000 0000 0000 0000 0000 0000  ................
	0000040: 0000 0000 0000 0000 0000 0000 0000       ..............

while the repaired file has a size of 78 (= 0x4e) bytes.  Some other files got
broken with random data.  I checked with debsums -c and it reported:

	root@cepheus:~# debsums -c
	>=������2[j�b��nw�������� in md5sums for irssi-scripts: ����gV{ڛ�
				 �N���L���Mg{����.�����`����ӈL���j$�kC1'� ��S� ���ݏ��
	debsums: invalid line (2) in md5sums for irssi-scripts: ����g�DPH��     숐���}�]g�����N�ci�5�h
	�w�W{SZ��q��F_�sR�[���ie�A|��Sv��@�@��;�5�'#c��$��l%���� ��T���$�!d�B�y
	debsums: invalid line (3) in md5sums for irssi-scripts: �����-ċE��yq�/7đ>�������Ў������Vu����V
	�+ɋA�f��:%O��_l���}������}� ��1���ȴϘ��=?��&��������F���mT�trZ�
	���1���enO%.�YN��=�k��@����\{8ɔw�x����z��-P!g�j����QV9u������)�m���5�l�8l �Rk5�;M���R���
												�fx��O
	gѝ��;����٠�HYfrc��9�����u�q���Ox߀`����~_�ƃ2"J;�Q$vl?�{�=V������ �[��\�d��n�!�UH��Y�D��j2I���*�
	[�c��G�������[��h*���������2A��m&����������ޥGЉ�;R�0��̦���
	...
	
I don't know exactly, but I think the damaged file here was
/var/lib/dpkg/info/irssi.md5sums and debsums repaired it!?

In theory this should not happen with a journaled fs, does it?

This is a 2.6.19.3 kernel, unfortunately tainted by madwifi.
There was nothing logged in dmesg and/or syslog.

	root@cepheus:~# xfs_info /var
	meta-data=/dev/hda9              isize=256    agcount=8, agsize=91619 blks
		 =                       sectsz=512   attr=0
	data     =                       bsize=4096   blocks=732952, imaxpct=25
		 =                       sunit=0      swidth=0 blks, unwritten=1
	naming   =version 2              bsize=4096  
	log      =internal               bsize=4096   blocks=2560, version=1
		 =                       sectsz=512   sunit=0 blks
	realtime =none                   extsz=65536  blocks=0, rtextents=0

I have to shutdown that machine because of a power cut by my supplier, but
probably I will use a boot cd to bring it up again ...

Best regards
Uwe

-- 
Uwe Kleine-König

http://www.google.com/search?q=1+degree+celsius+in+kelvin

Re: Problems with XFS in a power failure

[Emmanuel Florac]

Le Tue, 8 May 2007 11:16:14 +0200 vous écriviez:

> In theory this should not happen with a journaled fs, does it?

That's the opposite. It's the expected behaviour. It's especially
important to garantee proper power when using journaling filesystems. 

-- 
--------------------------------------------------
Emmanuel Florac               www.intellique.com   
--------------------------------------------------

Re: Problems with XFS in a power failure

[Federico Sevilla III]

On Tue, May 08, 2007 at 11:16:14AM +0200, Uwe Kleine-König wrote:
> my machine suffered a power failure while doing a apt-get upgrade.

Uh-oh. You hit the (in)famous binary nulls "issue". You may want to read
the FAQ entry: http://oss.sgi.com/projects/xfs/faq.html#nulls.

-- 
Federico Sevilla III
F S 3 Consulting Inc.
http://www.fs3.ph