August 30th, 2011, 12:52 PM
Incremental only vtapes
I am still new to Amanda and I have a large (mostly static) dataset to backup, 16TB on a RAID6 with growth of 30GB-40GB every week.
Right now I have a 25TB RAID6 server that I will be using in the short to medium turn with vtapes. I understand the importance of offsite backups, but the budget isn't there for tapes or storage. So I am using the hardware I have on hand in the meantime.
I am considering doing incremental only backups with 10GB vtapes once a week. with a runtape limit of 5 tapes.
This is outside the normal "let amanda do her own thing" usage. I reason this may be appropriate because of the static nature of the dataset (mostly growth, only a few updates of files). Am I setting myself up for failure later on? Or is this an okay method?
February 18th, 2012, 11:43 AM
(I realize this is an old post. Just thought it should at least show a response.)
You can't really do incremental backups without having a full backup. Nor would you want to. If you were just backing up the things that had changed, and the array failed, you would lose all the things that had not recently changed. You need the full backup as a reference against which to run the incremental backups, and you need the full backup to have full protection against losses or failures.
I understand the reasoning in trying to come up with this kind of strategy. The full backup of 16TB can take a long time. However, the Amanda approach would be to try to logically break up the 16TB into a fair number (say, 24) of Disk List Entries (DLEs) corresponding to different directories or subdirectories in the array. Then provide enough vtapes and a long enough dump cycle that the backups each day would run in a reasonable time. You'd have to play with these parameters a bit, but you would end up with something that would run full backups of a small number of DLEs together with incrementals of the bulk of the DLEs each day. It would smooth out the load over the dump cycle, which might be a couple of weeks of daily backups.
You may run into some issues because your 25TB array only has room for a single full backup all the contents of the 16TB array. That will get worse over time as more data is added. But you can juggle parameters until you come up with the best arrangement under the circumstances.
Tags for this Thread