ADSM-L

Re: Question on backup stgpool

2000-07-08 11:27:20
Subject: Re: Question on backup stgpool
From: "Slaughter, Bill" <BillSlaughter AT TUPPERWARE DOT COM>
Date: Sat, 8 Jul 2000 11:27:20 -0400
In my experience, I have found the same problems/features.

My resolution was to start multiple backup stgp commands. I actually start
them when the backup is still running (head start). I run the 2nd and 3rd
pass when I know the backups are complete. I also run all of them again
after migration from the primary tape to the offsite tape to catch anything
that was missed.

1. It appears that since the node/data is still active that some of my data
does not make it to offsite tape.

2. I wrote a shell script to issue "q content VVVVVV copied=no" for each
volume in the disk storage pools. It it determines that all data has been
copied it lets the Offsite process continue. The script also allows for an
8:00 cutoff by checking for the existance of a file
"time_for_offsite_process". If it sees this file I echo out the "q content"
output to list the data that did not get copied. One side benifet of this
process was that I found a Disk storage pool that was not being copied in
the morning processing.

3. backup stgp uses tape drive by node, I have found that even if you
specify maxprocess=2 it will only use one process per node in the storage
pool. I have a similar imbalance in that one node is over 120GB of Oracle
Data and using the above method we get all of our data out every day on
time.

When my window started getting tighter I got back almost an hour by running
the FULL backup earlier in the night and an incremental in the morning
process.


> -----Original Message-----
> From: Eric Tang [SMTP:tangeric AT HK1.IBM DOT COM]
> Sent: Friday, June 30, 2000 7:53 AM
> To:   ADSM-L AT VM.MARIST DOT EDU
> Subject:      Question on backup stgpool
>
> Hi All,
>
> Below is the timing of daily offsite processing this morning:
>
> Client NODE_A
> 06:00 to 07:30 Client backup of NODE_A (oracle RMAN/ConnectAgent) to disk
> storage pool
>
> Client NODE_B
> 22:00 to 04:44 Client backup of NODE_B (oracle EBU/ConnectAgent) to disk
> storage pool
>
> ADSM Server
> 04:30 to 07:57 Backup disk storage pool to copy storage pool
> 07:58 to 09:20 ADSM DB full backup
> 09:21 to  09:44     Move DRM
> .....
> .....
> 11:00 to 12:07 Backup disk storage pool to copy storage pool (2nd run)
> 12:08 to 15:26 Disk to tape stgpool Migration
> .....
>
> Question:
> 1. Found that files of the NODE_A are included in the 2nd backup stgpool
> but not in the 04:30 to 07:57 run. What is the logic of backup stgpool ?
> Are the files to be processed determined once backup stgpool command is
> issued? And that's why none of the files from NODE_A are included in the
> 04:30 run.
>
> 2. Again what is the logic of backup stgpool so that I can be sure that
> all
> files on NODE_B are offsited during the 09:00 move drm everyday.
>
> 3. The 04:30 backup disk stgpool is run with maxprocess=2, everyday. One
> of
> the process backup 16GB data and the other 62GB. This inbalance prolonged
> elapsed time as a whole. In terms of data volume, NODE_B contribute the
> most (~50GB), why didn't the 2nd process only backup the 50GB data to make
> the 2 processes more balance? Is it because of NODE_B's data is still
> going
> to ADSM server when backup stgpool started?
>
> Regards,
> Eric Tang
<Prev in Thread] Current Thread [Next in Thread>