January 17, 2003 at 9:20 am
Hello ALL:
I have to put together a plan for our database Which runs 24 X 7. It is transactional and Real time. We do not have replication or any clustered mechanism. WE are just starting off with limited funds.
Appreciate if someone could help us formulate the design for backup db and log so that in the event of a crash our loss is minimum.
Thanks
January 17, 2003 at 12:01 pm
I admin two MSSQL Server 2000 database systems that are 24x7 mission critical. We have a Dell 136T tape library with 24 tapes on one and a HP Superstore DLT vs80 single tape drive on the other. Both tape systems have Veritas Backup Exec 8.6 loaded.
My backup method is:
All backups are initially done using the SQL Server backup method, which creates a *.bak file. I use Veritas to copy the *.bak files to tape.
All tape copies are started 30 minutes after the SQL backup starts, with the exception of the Full Backup which starts one hour later.
I do one full backup every day at midnight, a differential every 4 hours except for the hour that I do a full backup, and a transaction log backup every hour except for the hours that I am doing a full or differential backup.
My transaction log backups are done to different *.bak files (trans1.bak, trans2.bak, trans3.bak). This way I should be able to do all of my restores from the *.bak files on the hard drive. But if I lose the hard drives or the *.bak files are corrupt, I have the tape copies. Also, I can store the tape copies off site as needed. Lastly, as I read one poster say on another site, in a real emergency (fire, etc) I can always grab the tapes out of my library and drives and run out of the building with them (yes I do keep a bag handy for this).
-Bill
January 17, 2003 at 12:34 pm
Similar to Bill here. Except we run full once a week, diffs the other nights. On some servers we run multiple diffs (up to 4) per day. In between we run tran logs every 15 minutes. Logs and diffs are "pushed" to a remote server since the tape system only runs at night. Gives me 2 copies of everything to keep data loss to 15 minutes.
Steve Jones
January 17, 2003 at 9:24 pm
Quite Simillar to the above. The only change being we push our backups to tape immediately (Add a step in the job).
January 23, 2003 at 3:03 pm
We have some servers that are in a cluster like environment (Marathon Endurance Server) but most of them are not. So what I do is I backup to a local disk for speed then copy ,another job step, across the LAN to the more stable server.
Then each night I use Veritas Backup Exec to copy from the stable server to tape and delete any backup files older then 24hrs.
I think copying the backup to another machines disks is worth doing because of the speed of retrieval compared to tape.
Nigel Moore
======================
January 23, 2003 at 3:14 pm
I originally did backup to tape Full once a week then diff other days. Then it wss moved to full everyday. But the problem was the same I was running into heavy production time for various reasons (17hours is too long). So am moving to local file backup (around 30mintues tops) then I will push to tape. Ours is a Quantum DLT7000 if I remember right. ANd I have a second at my disposal so I may do in combo. Some of the larger DBs have other recovery sources so we do them only once a week so I may put a weekly on tape then the others will be file then to other tape drive.
So I can keep the backups to the D drive. Otherwise I cut into the RAID array for the data drive.
Edited by - antares686 on 01/23/2003 3:14:31 PM
Viewing 6 posts - 1 through 5 (of 5 total)
You must be logged in to reply to this topic. Login to reply