Backup fials due to zipping process

[pa8700]

Hi ,


I have a file system for oracle archive logs.
We have a cron job to zipp the new files once in 15 minitus.We are using DP to take the backup.WE are failingto take the files, which are being used by the zipping process at the time of backup.If we disable the cron job for the backup period the filesystems will got full.

Can some body sugesst a solution.

[pa8700]

wether simple filesystem snapshot can solve my issue.

[Tingli]

Which process affects which one? (zip and backup)

[TTr]

> If we disable the cron job for the backup period the filesystems will got full

How much free disk space do you have and how long does the DP job take?

The filesystem snapshot requires disk space which in your case you don't seem to have. The disk space requirement for the snapshot can be very high since you are saying you are running zip jobs and have a lot of file i/o.

You should try and run the zip and DP jobs in sequense so that DP starts after zip is finished. Also the DP job should be limited only to the archive log volume so that it ends quickly and you can restart the zipping process.

If you have a lot of archive logs, you need more disk space and that's that.

[pa8700]

I am not concerned about the disk space.I can create an equal size lv and assign to use for snapshot.

My query is that wether Data protetor can be used on a online snapped fs.

[Mel Burslan]

If you know what you are doing, what you said is very much possible.

Few minutes/hours from the backup, create a mirror for the logical volume. A minute or two before the backup, break the mirror and export the physical volumes using map files. Create a new volume group on the fly and import these exported physical volumes. Mount the LV's on this new VG onto some dummy mountpoint. Using DP, get the backups of this dummy mountpoint. Most probably, before starting the backup, you will want to delete the incomplete zip files but this should not be a big problem. We use this method backing up our databases daily. It need a bit of planning but once you get it going, it runs w/o any major problems.

Also, as a caveat of this method, whne and if you need to restore, you may need to restore to the dummy mountpoint and move the files to their original locations which will add another level of complexity to your operations.

In a slightly unrelated note, if creating a new LV to take a snap of this volume is not a problem, why don't you double the size of the volume then stop zipping before the backup starts and let the newly expanded space fill up while your backup runs then turn on the zipping process when backup finishes ? Just an idea...

On a friendly note, consider increasing your points assignment ratio if you wish to get more responses to your questions.

[TTr]

> If we disable the cron job for the backup period the filesystems will got full.

>I am not concerned about the disk space.I can create an equal size lv and assign to use for snapshot.

Then I fail to see the problem. If disk space is not the issue why not extend the archive log volume? How busy is your database to be filling up the archive log volume if you just turn off the zipping process? How long does it take for DP to backup the archive volume? Maybe you should have a dedicated DP job for only the archive volume.

[pa8700]

see I have 54 arch filesystem for 54 Databases on a single system.So time calculations are bit complex.

VxFS snap is a readonly, so I want to know wether DP is able to take the backup .If DP is not updating any inodes it should.