cancel
Showing results for 
Search instead for 
Did you mean: 

How to remove a QIO device file on solaris?

Hi Experts,

A device file was created as a result of running the Veritas qiomkfile command:

# cd /sybase/k2rpt_sql/TOR_K2RPT_UAT/sybase_devices
# /opt/VRTS/bin/qiomkfile -s 2000m log06.dat.qio
# ls -l log06.dat.qio
lrwxrwxrwx   1 root     root          26 Sep 18 10:29 log06.dat.qio -> .log06.dat.qio::cdev:vxfs:
# ls -l .log06.dat.qio::cdev:vxfs:
crw-r--r--   1 root     root     283, 35 Sep 18 10:29 .log06.dat.qio::cdev:vxfs:

 

Now I want to remove the link, which I have no problem doing so with rm command,  and the device file, on which rm command has no effect:

 

# rm -f .log06.dat.qio::cdev:vxfs:
# ls -l .log06.dat.qio::cdev:vxfs:
crw-r--r--   1 sybase   sybase   283, 35 Sep 18 10:29 .log06.dat.qio::cdev:vxfs:
# rm -f ".log06.dat.qio::cdev:vxfs:"
# ls -l .log06.dat.qio::cdev:vxfs:
crw-r--r--   1 sybase   sybase   283, 35 Sep 18 10:29 .log06.dat.qio::cdev:vxfs:

Any help is appreciated.

Alex

2 Solutions

Accepted Solutions
Accepted Solution!

The SF Oracle guide say:   To

The SF Oracle guide say:

 

To disable Quick I/O
 
1 If the database is running, shut it down.
 
2 To change Quick I/O files back to regular VxFS files, run the following
 command from the directory containing the mkqio.dat list:
$ /opt/VRTSdbed/bin/qio_convertdbfiles -u
The list of Quick I/O files in the mkqio.dat file is displayed. For example:
.file1::cdev:vxfs: --> file1
.file2::cdev:vxfs: --> file2
.file3::cdev:vxfs: --> file3
.file4::cdev:vxfs: --> file4
.file5::cdev:vxfs: --> file5
 
The qio_convertdbfiles command with the undo option (-u) renames the
files from .filename to filename and removes the symbolic link to .filename
that was created along with the Quick I/O files.
 
3 To remount the file system with Quick I/O disabled, use the mount -o noqio
command as follows:
# /opt/VRTS/bin/mount -F vxfs -o remount,noqio /mount_point
 
So I guess you need to find or create mkqio.dat file so that it contains file you want to delete to convert Quick IO file to normal file and then you can delete the normal file.
 
 
Mike

 

View solution in original post

Accepted Solution!

You MAY be able to

You MAY be able to edit mkqio.dat to contain just the file you want (making a copy of file first) and then restore mkqio.dat back afterwards.

Mike

View solution in original post

4 Replies
Accepted Solution!

The SF Oracle guide say:   To

The SF Oracle guide say:

 

To disable Quick I/O
 
1 If the database is running, shut it down.
 
2 To change Quick I/O files back to regular VxFS files, run the following
 command from the directory containing the mkqio.dat list:
$ /opt/VRTSdbed/bin/qio_convertdbfiles -u
The list of Quick I/O files in the mkqio.dat file is displayed. For example:
.file1::cdev:vxfs: --> file1
.file2::cdev:vxfs: --> file2
.file3::cdev:vxfs: --> file3
.file4::cdev:vxfs: --> file4
.file5::cdev:vxfs: --> file5
 
The qio_convertdbfiles command with the undo option (-u) renames the
files from .filename to filename and removes the symbolic link to .filename
that was created along with the Quick I/O files.
 
3 To remount the file system with Quick I/O disabled, use the mount -o noqio
command as follows:
# /opt/VRTS/bin/mount -F vxfs -o remount,noqio /mount_point
 
So I guess you need to find or create mkqio.dat file so that it contains file you want to delete to convert Quick IO file to normal file and then you can delete the normal file.
 
 
Mike

 

View solution in original post

Hi Mike, Thanks for your

Hi Mike,

Thanks for your valuable information. I learned a new command!

However, the command will convert all QIO files back to regular files, which is not desirable. My senario is that in case of a QIO file was added with wrong size/filename, I want to delete it and recreate it. Is there a way we can do this for a single file?

Best regards,

 

Alex

Accepted Solution!

You MAY be able to

You MAY be able to edit mkqio.dat to contain just the file you want (making a copy of file first) and then restore mkqio.dat back afterwards.

Mike

View solution in original post

Micke, Thank you

Micke, Thank you again.

 

Alex