-
We are backing up data from a document management system that is based on Oracle.
Metadata is stored in Oracle. When TIFF images and other files are checked in, the actual files are stored on a secure volume outside Oracle while only references to the files are kept in Oracle. The files are not stored as blobs.
We now do incremental hot backups of Oracle. We'd like to also backup up the newly checked in files as part of the Oracle backup. (Actually we now are doing two backups, but the two can't be guaranteed to be in sync with each other.) How can we find out from the ARCHIVE LOG data the references to the newly checked in files?
With Oracle 8i we might be able to use Log Miner to parse out the references inserted into an Oracle table.
Is that a good approach?
Right now we're using Oracle 8.0.6. Is there a similar tool available for this older version of Oracle?
Thanks
-
how do you reference those TIFF? With bfile?
When you say you do two backups you mean you do incremental hotbackup in Oracle and just simple file backup with TIFFs?
Oracle always provide consistency, if you get 3 TIFF during your hotbackup then those new insertions will be recorded on logs and the datafiles, after hotbackup of the tablespace ends there will be a checkpoint which syncronize datafile header and fatafile blocks.
I am not sure if this is what you are wondering
-
I'm not quite sure what you mean either. Do you mean that at backup time, you want to find a list of all the newly checked in files?
Couldn't you use a database field to compare dates, or a flag for new, or something of that nature? Or let your check in/check out program set some information?
-
We don't have a lot of control over changing how the doc management system works by adding flags to be set.
Yes there is an incremental hotbackup in Oracle and just simple file backup with TIFFs.
If we backup with ARCHIVELOG turned on, then I think we should be able to write something to parse the log file to pull out the references to the TIFF files and then back up only those files.
We are using Oracle 8.0.5. I understand the Log Miner might be able to help us parse the log files. But Log Miner is in 8i.
What's the simplest way to pull the TIFF references from the Log file after the Oracle backup?
-
Could you tell me more about: "after hotbackup of the tablespace ends there will be a checkpoint which syncronize datafile header and fatafile blocks".
How might we use these to automate our TIFF file backup?
Thanks
-
To find the files that have been changed, you can use the unix find command.
find -ctime +
When you remember the last date you made a backup off your files, you can calculate the needed in your search.
Some options for the find command :
-atime Number Evaluates to the value True if the file has been accessed in
Number-1 to Number multiples of 24 hours. For example, -atime 2 is true if the
file has been accessed within 24 to 48 hours.
-ctime Number Evaluates to the value True if the file i-node (status
information) has been changed in the specified number of 24-hour periods.
-mtime Number Evaluates to the value True if the file has been modified in
Number-1 to Number multiples of 24 hours.
Hope this helps
Gert
-
Some characters are gone on the previous post.
new example of find command :
find /home/tiffdata -ctime +1
Gert
-
Actually we're doing something like that for the current backup of the TIFF document files.
This is close, but because we're doing hot backups, I don't think we can guarantee an exact match of one set of TIFF document backups to the Oracle data. People can be checking documents in while the Oracle backup is running. So depending how we choose the timestamp, we might miss a file or pick up a file that should be part of the next backup.
-
During a hotbackup of a tablespace the datafiles that belong to that tablespace can still be accessed and all changes during the hotbackup reflects in the datafiles, as long as those "pointers" that points the TIFF are present and the TIFF are in the place where they should you wont get an consistency. I see you mentions incremental backup do you do this with RMAN?
-
We are not using RMan yet.
We're using DBBreeze for the backups.
Sorry, but I'm still not clear. First we want to do an Oracle hot backup.
Then we'd like to do a 'backup' or copy of all the newly checked in TIFF files. This TIFF backup should only contain the newly checked in files -- only the files referenced by the new rows added to a table that was part of the Oracle hot backup. How can I identify only those TIFF files or how can I find only those TIFF file pointers so we know which TIFF files to back up?
I want to be able to get the TIFF location information from the Hot backup data. How do I access/read the hot backup data to find out the new rows added to an Oracle table?
Someone suggested that we use the UNIX command to find the date/time stamp on the TIFF files to find the new ones. This is what we are doing now.
But we want to maintain a series of files associated with the Oracle hot backups and for each hot backup we also want to have a set of TIFF files.
Thanks
-
Without logminer (which I haven't used and don't know if that would really help anyway), I really don't think there is a way to grab that information from the log files.
You mentioned you don't have much freedom to change the doc management system, but if you really require this functionality you might have to. A simple date column saying when the file was checked in would suffice. Actually you could probablly do it without the date field, if the OS timestamp is the same as when the file was checked in.
One thing you'll need though is a log saying when the backups were done so you can create a series, and know when the previous backup was done so you can only take files newer then the latest backup.
For a basic example, say you take your first backup on Day 1.
For your second backup, on Day 5, every new file will have a date greater then Day 1. Copy only those files.
On your third backup, Day 10, you know to only take files with dates greater then Day 5.
It would be easiest with a date column inside Oracle, but if you really can't add that column you could do it with OS timestamps.
[Edited by pwoneill on 11-27-2000 at 11:22 AM]
