cp, cat, mv error on .dbf file

[Joe Profaizer]

We are trying to mv this file to another filesystem. Below is an explanation from our DBAs and as far as we know there is no problem accessing the database.

Here's the error we're getting when trying to read this file. We get this on a cp, cat, mv or any command which reads the file. The file does exist and occurs after getting quite a ways through the file:

rga8:/prd01/stl_rdvl >> cat users01.dbf > /dev/null
cat: read error: No such file or directory

rga8:/prd01/stl_rdvl >> ls -l users01.dbf
-rw-rw---- 1 oracle dba 2097160192 Aug 28 14:37 users01.dbf


Here's the output from tusc:

read(3, 0x6fff09a0, 8192) ................................ ERR#6 ENXIO

Please assist.

[Stefan Farrelly]

Are you absolutely sure no process has the database file open at the time your trying to copy it ? (fuser reports no processes).

Have you tried using dd to copy it;

dd if= of= bs=64k

Is the file over 2 GB and the new filesystem youre copying it to setup to handle largefiles > 2 GB ?

[Joe Profaizer]

I just found out that there are 6 files that have this problem. We made sure there was no other access to the files (fuser). After trying the dd command we received the following error:

no such device or address

The file is less than 2GB and largefiles are supported on source and target mountpoints.

..Joe

[Christopher McCray_1]

Does your syslog say anything reference this; perhaps inode errors? Just a thought.

Chris

[harry d brown jr]

What user are you logged on as, and does that user have access to the group "dba"???

[Joe Profaizer]

It failed on dd and we just tried to back the files up via Omniback and received virtually the same errors. We logged in as root to try this also.

Our DBA is going to attempt to use RMAN to back up the tablespaces and then try a recovery.

[Stefan Farrelly]

Hmm, cat coming up with a read error, and dd reporting no such device or address implies that they underlying hardware your file is sitting on has a problem.

When dd came up with its error how many blocks through was it ? It should say something like; X blocks in, X blocks out, then the error.

Can you also try this in increments;

dd if= of=/dev/null bs=1024k count=1

This will read the first 1 MB of the file. Did it work ?
Then slowly increase the count value until you encounter the error.

[Bill Thorsteinson]

Are you sure the database isn't usng raw devices?
Some DBAs choose to use them to get faster access.
This is true on some systems, but causes management
problems such as this.

The 'dd' command should be able to move the files.
It may spit out an error on end of file, but should
specify the number of blocks copied. Use the
bs=1024 option on the dd command and compare
the blocks coped to the value du -k gives for the
database file being copied.