How incremental backups work on UNIX?

[NB-OPS]

Hi,

I understand to back up UNIX clients, NetBackup compares mtime of the files to see if an incremental backup is needed. Let's say file1 

- was modified on 1/1

- was backed up by a full backup on 1/2

- was modified on 1/3

- an incremental will be done on 1/4

During the incremental backup, does NetBackup compare 1/3 with 1/2? Where does it find the information about 1/2? from the catalog? or the last full backup start time, end time? 

Thanks,

[revarooo]

It finds the previous backup from catalog, when you did the full on 1/2 that information is all captured within that catalog and stored. In that case all files are backed up.

When inc backup fires on 1/4 it checks if the files has been modified since the last backup, in this case a file or files changed from 1/3 and compares with the backup of 1/2 to see if there is a change in the mtime, there is, so it backs it up. If there is no change it doesn't need to back it up until that file/s change or until the next full backup runs.

[NB-OPS]

Thanks for the quick reply.

So let's say the full backup took 18 hours (2:00-20:00 on 1/2) and this particular file was backed up at 6:00 on 1/2. Does NetBackup actually compare the current mtime of the file with 6:00 1/2? It seems to be a lot of comparisons to be made if I have millions of files. 

Thanks again!

[revarooo]

It logs that the file was backed up at completion of the full (if memory servers correctly).

If you have millions of files changing regularly enough to be caught by incremental backups you would be better using Flashbackup to backup at a block level rather than file level as backing up millions of files even with a full backup can be slow, flashbackup would be quicker (though restores can be slower)

[NB-OPS]

> It logs that the file was backed up at completion of the full

I suppose you meant at the completion of the full backup of that particular file (not the end of the full backup job)?

I drilled down to the hours in my previous update because if the file is backed up at 6:00 on 1/2 and the full backup job does not finish until 20:00 on 1/2,  in between 6:00 and 20:00, this particular file could have been modified. In this case, unless NetBackup record the individual file backup time after each and every file is backed up, it will miss this file in the next incremental. So I guess NetBackup records in the catalog the timestamp when each and every file is backed up. 

It seems to have alot of to write to the catalog and alot of comparisons to make for incrementals backups..

Thanks!

[Yasuhisa_Ishika · Accepted solution]

NetBackup compares mtime of each file with backup time of previos backup - backup time is start time of backup operation(=when backup go active). In this case, NetBackup compares mtime with 06:00 on 1/2.
Backup elapsed time does not affect incremental backups coverage.