file shrank by 116557 bytes - why?

[Eric]

Every time i use tar to create a backup this 1 particular file does this:

/bin/tar: usr/lib/sane/libsane-umax.so.1.0.12: File shrank by 116557 bytes;
padding with zeros

Is that some kind of special file? What is happening here? It acts like
reading the file makes it change size.

Thanks
Eric

[Joe Beanfish]

Tar's probably complaining that it read less than stat said was there.
Check it out with
ls -l usr/lib/sane/libsane-umax.so.1.0.12
wc usr/lib/sane/libsane-umax.so.1.0.12

Check /var/log/messages for system problems.

[Jeroen Geilman]

Eric wrote:
> Every time i use tar to create a backup this 1 particular file does this:
>
> /bin/tar: usr/lib/sane/libsane-umax.so.1.0.12: File shrank by 116557 bytes;
> padding with zeros

It's a sparse file, i.e. one with extra space in it reserved for future
writing, or because there needed to be data later on in the file but not
in the middle.
Weird concept if you come from DOS, but really quite efficient.

> Is that some kind of special file? What is happening here? It acts like
> reading the file makes it change size.

Not exactly - but *copying* it makes tar realise it is a sparse file so
it just skips those bits.

Try tar with and without --sparse and see the difference.

--
J

Where does the shit go ?

[Hans J. Heerwig]

Hi, I have the same problem and so far I have not solved it.

"Eric" <notV...@comcast.net> schrieb im Newsbeitrag
news:0Wjhd.41420$R05.16355@attbi_s53...

Descrition:

1. Running sles8 (SuSE) Linux with kernel 2.4.19, filesystem used: ext3

2. System is an oracle Release 9.2.0.5.0 server

3. backup with tar czf /dev/rmt0 xxx yyy zzz
(xxx yyy zzz = directories)
oracle complete shutdown with "shutdown immediate" with sqlplus statement
(the output of shutdown oracle proceedure and timestamps is filed in a
generated logfile)

4. error in mail (root) because it runs as cron root:
tar: ora_dat/xxx/yyy_IDX_00.dbf: File shrank by 594493952 bytes; padding
with zeros

tar: ora_dat/xxx/yyy_123_00.dbf: File shrank by 5221016576 bytes; padding
with zeros

tar: Error exit delayed from previous errors

The file size is 15,x (idx) and 5,x (123) GB

5./var/log/messages:
Oct 28 23:48:01 becker-ora kernel: scsi0: ERROR on channel 0, id 0, lun 0,
CDB:

0x28 00 00 7c 98 22 00 04 00 00

Oct 28 23:48:01 becker-ora kernel: Info fld=0x7c9a78, Current sd08:03: sns =
f0

3

Oct 28 23:48:01 becker-ora kernel: ASC=11 ASCQ= 0

Oct 28 23:48:01 becker-ora kernel: Raw sense data:0xf0 0x00 0x03 0x00 0x7c
0x9a

0x78 0x0a 0x22 0x43 0x05 0xd0 0x11 0x00 0x00 0x80 0x00 0x81

Oct 28 23:48:01 becker-ora kernel: I/O error: dev 08:03, sector 3892712

Oct 28 23:48:01 becker-ora kernel: klogd 1.4.1, ---------- state
change --------

--

Oct 28 23:48:01 becker-ora kernel: Inspecting
/boot/System.map-2.4.19-64GB-SMP

Oct 28 23:48:05 becker-ora kernel: Loaded 21170 symbols from
/boot/System.map-2.

4.19-64GB-SMP.

Oct 28 23:48:05 becker-ora kernel: Symbols match kernel version 2.4.19.

Oct 28 23:48:05 becker-ora kernel: Loaded 664 symbols from 18 modules.

Oct 28 23:48:05 becker-ora kernel: scsi0: ERROR on channel 0, id 0, lun 0,
CDB:

0x28 00 00 7c 9a 72 00 00 08 00

Oct 28 23:48:05 becker-ora kernel: Info fld=0x7c9a78, Current sd08:03: sns =
f0

3

Oct 28 23:48:05 becker-ora kernel: ASC=11 ASCQ= 0

Oct 28 23:48:05 becker-ora kernel: Raw sense data:0xf0 0x00 0x03 0x00 0x7c
0x9a

0x78 0x0a 0x22 0x43 0x05 0xd0 0x11 0x00 0x00 0x80 0x00 0x81

Oct 28 23:48:05 becker-ora kernel: I/O error: dev 08:03, sector 3892712

Oct 28 23:48:20 becker-ora kernel: scsi0: ERROR on channel 0, id 0, lun 0,
CDB:

0x28 00 00 8f af e2 00 04 00 00

Oct 28 23:48:20 becker-ora kernel: Info fld=0x8fb0a6, Current sd08:03: sns =
f0

3

Oct 28 23:48:20 becker-ora kernel: ASC=11 ASCQ= 0

Oct 28 23:48:20 becker-ora kernel: Raw sense data:0xf0 0x00 0x03 0x00 0x8f
0xb0

0xa6 0x0a 0x22 0x43 0x06 0xb4 0x11 0x00 0x00 0x80 0x00 0x81

Oct 28 23:48:20 becker-ora kernel: I/O error: dev 08:03, sector 5143576

Oct 28 23:48:25 becker-ora kernel: scsi0: ERROR on channel 0, id 0, lun 0,
CDB:

0x28 00 00 8f b0 a2 00 00 08 00

Oct 28 23:48:25 becker-ora kernel: Info fld=0x8fb0a6, Current sd08:03: sns =
f0

3

Oct 28 23:48:25 becker-ora kernel: ASC=11 ASCQ= 0

Oct 28 23:48:25 becker-ora kernel: Raw sense data:0xf0 0x00 0x03 0x00 0x8f
0xb0

0xa6 0x0a 0x22 0x43 0x06 0xb4 0x11 0x00 0x00 0x80 0x00 0x81

Oct 28 23:48:25 becker-ora kernel: I/O error: dev 08:03, sector 5143576

---------

so far the facts.

I got the same prob earlier on another system (suse 7.2 / oracle 8.1.7).

In all casses the files contains a lot of space for future use of the db.

Is was just a testsystem and it was not possible to investigate it (system
needed for other usage / new installation)

I dont' beleave the messages (scsi dev. channel 0, id 0, etc. with error,).

The indicated files work without probs.

Tonight I' ll try to tar just one effected file (see above, the smaller one)
and I will see what happens.

Evtl. it may be just a capacity prob (DDS3 dat tape used ...) the timestamp
of mail and message is at the end of the tar job.

--------------

Perhaps (if wanted) i will report the result of further investigations.

On the other hand it would be very friendly to receive some tips or info.

greetings and the best wishes to all colleagues

Hans

[Hans J. Heerwig]

Further Tests:

Just tried to copy the relevant files.

Test cp 123:

becker-ora:/ora_dat/xxx # cp yyy_123_00.dbf /home/admin/test

cp: reading `yyy_123_00.dbf': Input/output error

becker-ora:/ora_dat/xxx #

var/log/messages

Nov 4 18:45:12 becker-ora kernel: scsi0: ERROR on channel 0, id 0, lun 0,
CDB:

0x28 00 00 8f b0 a2 00 00 08 00

Nov 4 18:45:12 becker-ora kernel: Info fld=0x8fb0a6, Current sd08:03: sns =
f0

3

Nov 4 18:45:12 becker-ora kernel: ASC=11 ASCQ= 0

Nov 4 18:45:12 becker-ora kernel: Raw sense data:0xf0 0x00 0x03 0x00 0x8f
0xb0

0xa6 0x0a 0x22 0x43 0x06 0xb4 0x11 0x00 0x00 0x80 0x00 0x81

Nov 4 18:45:12 becker-ora kernel: I/O error: dev 08:03, sector 5143576

Nov 4 18:45:12 becker-ora kernel: klogd 1.4.1, ---------- state
change --------

--

Nov 4 18:45:12 becker-ora kernel: Inspecting
/boot/System.map-2.4.19-64GB-SMP

Nov 4 18:45:12 becker-ora kernel: Loaded 21170 symbols from
/boot/System.map-2.

4.19-64GB-SMP.

Nov 4 18:45:12 becker-ora kernel: Symbols match kernel version 2.4.19.

Nov 4 18:45:12 becker-ora kernel: Loaded 664 symbols from 18 modules.

Source Directory:

-rw-r----- 1 oracle dba 5242888192 Nov 4 18:42 yyy_123_00.dbf

Target Directory:

-rw-r----- 1 root root 21868544 2004-11-04 18:45 yyy_123_00.dbf

Test cp idx:

becker-ora:/ora_dat/xxx # cp yyy_IDX_00.dbf /home/admin/test

cp: reading `yyy_IDX_00.dbf': Input/output error

var/log/messages

Nov 4 19:04:41 becker-ora kernel: scsi0: ERROR on channel 0, id 0, lun 0,
CDB:

0x28 00 00 7c 98 22 00 04 00 00

Nov 4 19:04:41 becker-ora kernel: Info fld=0x7c9a78, Current sd08:03: sns =
f0

3

Nov 4 19:04:41 becker-ora kernel: ASC=11 ASCQ= 0

Nov 4 19:04:41 becker-ora kernel: Raw sense data:0xf0 0x00 0x03 0x00 0x7c
0x9a

0x78 0x0a 0x22 0x43 0x05 0xd0 0x11 0x00 0x00 0x80 0x00 0x81

Nov 4 19:04:41 becker-ora kernel: I/O error: dev 08:03, sector 3892712

Nov 4 19:04:45 becker-ora kernel: scsi0: ERROR on channel 0, id 0, lun 0,
CDB:

0x28 00 00 7c 9a 72 00 00 08 00

Nov 4 19:04:45 becker-ora kernel: Info fld=0x7c9a78, Current sd08:03: sns =
f0

3

Nov 4 19:04:45 becker-ora kernel: ASC=11 ASCQ= 0

Nov 4 19:04:45 becker-ora kernel: Raw sense data:0xf0 0x00 0x03 0x00 0x7c
0x9a

0x78 0x0a 0x22 0x43 0x05 0xd0 0x11 0x00 0x00 0x80 0x00 0x81

Nov 4 19:04:45 becker-ora kernel: I/O error: dev 08:03, sector 3892712

Source Directory:

-rw-r----- 1 oracle dba 15728648192 Nov 4 18:42 yyy_IDX_00.dbf

Target Directory:

-rw-r----- 1 root root 15134150656 Nov 4 19:04 yyy_IDX_00.dbf

-------------

The Tests show, that there is really a problem witin the file/disk.

the "123" file is in the beginning of its lenght defective, the IDX at the
end.

What possibilities are there to "repair" the files? Who could help?

Many thanks in advance,

Hans

[Michael Buchenrieder]

"Hans J. Heerwig" <ha...@heerwig.com> writes:

[...]

>Oct 28 23:48:01 becker-ora kernel: ASC=11 ASCQ= 0

>Oct 28 23:48:01 becker-ora kernel: Raw sense data:0xf0 0x00 0x03 0x00 0x7c
>0x9a

>0x78 0x0a 0x22 0x43 0x05 0xd0 0x11 0x00 0x00 0x80 0x00 0x81

>Oct 28 23:48:01 becker-ora kernel: I/O error: dev 08:03, sector 3892712

[...]

Uh-oh.

Make a backup immediately. Your SCSI disk is reporting a read error
on sda3, which means that a drive failure is imminent, in case you'd
probably loose all data on that disk.

Michael

--
Michael Buchenrieder * mi...@scrum.greenie.muc.de * http://www.muc.de/~mibu
Lumber Cartel Unit #456 (TINLC) & Official Netscum
Note: If you want me to send you email, don't munge your address.