snapshot filesystem status

[Tom Reeve]

I would like to use a snapshot filesystem to backup a production environment. I have successfully used this method in test, but I am interested to find a way to check the status of a snapshot filesystem - specifically, how "full" it is, so I can see if I'm going to have a problem with a running backup. I've searched this site and available docs, and haven't found a way to do this. Anybody out there know a way to check this? Thanks for checking - - Tom

[James R. Ferguson]

Hi Tom:

'bdf' gives an excellent, terse summary of the utilization of mounted filesystems.

...JRF...

[Vincenzo Restuccia]

In /path try "bdf .".

[Thierry Poels_1]

hi,
I'm afraid that bdf on the snapshot show the same result as for the original filesystem.
to be continued ...
Thierry.

[Tom Reeve]

Indeed, bdf does that. However, not in this case. With a snapshot file system, bdf reports the same % used for the original and the snapshot filesystem. I probably should have been more explicit - here's the command I used to create the snapshot filesystem:

mount -F vxfs -o snapof=/org_vol /aux_vol /aux_dir

The concern is that during a period of higher db activity, the snapshot filesystem could become full, causing the backup to fail, with no backup possible until the following day.

Thanks for looking at this.

[A. Clay Stephenson]

Tom,
I too looked into this issue when I started snapshot ORACLE backups. I was never able to
devise a method of determining the current
usage of the snapshot buffer. I even explored
ioctl's. However, this only makes sense in that the whole purpose of the snapshot is to make an indistinguishable copy of the original.
From a practical point of view, I have used a buffer that is 20% of the ORACLE data; my backups take about 4 hours to complete writing to 2 DLT7000's. The 20% value has never caused a problem and a backup has never failed because of a full snapshot buffer. The total downtime including bouncing the database is under 2 minutes.
One gotcha I did find that could hang the system was a failure of the snapshot disk(s).
I put my snapshot buffers in a separate volume group but did not mirror the logical volumes. A disk failed and the system hung. I/O's to the original file system were suspended. To avoid this, mirror the snapshot logical volumes.