LinuxQuestions.org

LinuxQuestions.org (/questions/)
-   Linux - General (http://www.linuxquestions.org/questions/linux-general-1/)
-   -   an incremental backup strategy confusion (http://www.linuxquestions.org/questions/linux-general-1/an-incremental-backup-strategy-confusion-505514/)

prozac 11-28-2006 04:46 AM

an incremental backup strategy confusion
 
I am planning to backup 6-7 servers using bacula. I have been considering a backup plan whereby I can be "able to access/restore the old files on daily, weekly and monthly basis for 3 months." By which I mean I should be able to access/restore data to any point within this 3 month.

I have decided to go with the GFS scheme on this. My current data-size to backup is around 80 G in total. I expect (i am still confirming in this) a daily increment of 0.5 M and a weekly increment of nearly 30 M.

With the above parameters I have tried to calculate the total size of data I would be backing up for these 3 months.

80Gx3(months)=240G (A full backup every month)
30Mx4(weeks)x3(months)=360M (A weekly incremental backup)
0.5Mx5(days)x4(weeks)x3(months)=30M (Daily incremental backup)
---------------------------
Total data size~=241G
---------------------------

With this strategy, I think If I required to restore to Nov 7 2006 then I would first restore the monthly backup for the month of Oct/Nov. then 1st week incremental backup for the same month and then the daily incremental backup from that day till I reach Nov 7.

I am confused. Is this how it works? Or, Will the above strategy work? I am confused on the weekly incremental section. Am I supposed to take a full backup each week or just an incremental backup?

Any useful comments is appreciated. Basically I am just confused.

TIA,
prozac.

MensaWater 11-28-2006 02:17 PM

Haven't used Bacula but based on the information on its site and other backup products/tools I've used the answer is yes it will work.

According to their definition:
Quote:

Incremental
A backup that includes all files changed since the last Full, Differential, or Incremental backup started. It is normally specified on the Level directive within the Job resource definition, or in a Schedule resource.
I will say that most people don't do it that way (at least for larger environments). Typically they'll do the full on a regular basis (once a week) then do a what bacula calls a differential (backs up everything changed since the last full) on a daily basis. This would make restores a lot faster because you'd only have to restore the most recent full and the most recent differential.

In your scenario you'd have to restore the most recent full and each subsequent incremental. (Not just the weekly incremental but each day's incremental.) "Level" is a keyword in the above quote. If they allow you to do "Levels" as some utilities do and you set your full to highest level then your weekly to next lower level then your dailies to next lower level you can probably do it the way you said. I didn't read their documentation to see if that's how they do levels but its pretty standard in backup solutions.

prozac 11-28-2006 10:25 PM

Thanks once again for the response. Yes, my backup plan in practice would mean longer restore times and from many disks (or volumes). But the disk size and cost is reduced which is a factor I needed to consider. One thing that still confuses me-just say I did a full backup each week instead of the incremental and keep weekly backups for 3 months, that would mean:

80Gx3(months)=240G (A full backup every month)
80Gx4(weeks)x3(months)=960G (A weekly incremental backup)
0.5Mx5(days)x4(weeks)x3(months)=30M (Daily incremental backup)
---------------------------
Total data size~=1200G
---------------------------

Now when I need to do a full restore, where do I start from? Do I start from restoring Oct/Nov monthly backup then the 1st week full backup and then the incremental ones till I reach Nov 7? I am confused, Look at the sheer size of doing restores, First nearly 80G of monthly full backup, again nearly 80G of the weekly backup and then some MB's of daily incremental backup? Or, start from the weekly one? Which seems the way to do but then how do I justify the monthly backups (I mean when to use them anyway?).

Again TIA,
prozac.

prozac 11-28-2006 10:29 PM

One more thing though, I didn't find anything related to compression on bacula documentation. I know compression ratio largely depends on the data but since most of the data I would be backing up are databases and simple text files, It would be a help to know in advance the total approx. disk size I would be needing. I've read bacula uses zlib as it compression engine, that would mean the compression provided would roughly be equal to zip compression. What do you think?

trickykid 11-29-2006 06:31 AM

Industry standard average is around 3% change or addition every day that most backup and storage solution enterprises will use to calculate incremental changes to assist in predicting the amount of data to backup.

prozac 11-29-2006 10:39 PM

Thanks for all the response. I am setting up a test-bed to experiment w/ the plan and I will surely post my findings, whatever and however it be. Meanwhile, I have realized that I was not confused. I was just not confident. Thanks for helping me out.


All times are GMT -5. The time now is 12:39 AM.