Recommendation on tapes and tape job

[burak]

Hello,

I need a little bit help how to handle the backup to tape jobs with my tapes. Our original idea was to use 20 tapes, 12 for months, 4 for weeks and 4 tapes for years (we bought a 20 pieces package), so we already labeled those tapes with January to December, Week 1 to Week 4 and Year 2017 to 2020. Monthy and weekly tapes should rotate and yearly tapes should never be overwrite.

Now, I am confused about the media sets, the retention and the rotation of those tapes. Since, my repository has a size of 3,2TB and we use LTO-6 tapes, we can only save 2,5TB data on the tapes. This means, I can't save the complete repository on the tape. We have 3 backup jobs with active-full on each sunday and we want to switch the tape only once in a week for example each Monday. A full backup of all jobs together is only 300GB big.

I think, GFS media pool and backup job would be a good choice here and I read in the documentation that a GFS job always store a full-backup on the tape and since I use active-full, it does not create a virtual full backup and only copy the last full backup in my repository. However, if I look at this picture here



my tape "January" would be useless because the yearly-tape is also the tape for January?

If I use the GFS job, is the tape always overwritten with the newest full backup or is the newest full backup attached to the backups on tape? If the tape is full, does it just rewrite from the start? With 300GB backup size, I could save almost 8 full backups on a tape.

Would be glad if someone can give me some recommendation for my scenario.

Thank you!

[lyapkost]

Hi Burak. Your plan looks nice. There's no need to compare the entire size of your repository and LTO6 tape, only the size of full backups counts which is in your case much less than a tape may store.

my tape "January" would be useless because the yearly-tape is also the tape for January?

If the schedule for backup periods overlap, only the backup for the media set with the longest retention will be archived (UG). In case you want to have all tapes written, change the schedule so that gfs dates of different media sets do not overlap, for example, first sunday for yearly and second sunday for monthly media sets.

If I use the GFS job, is the tape always overwritten with the newest full backup or is the newest full backup attached to the backups on tape?

It depends whether 'append' option is enabled or not. By default it is disabled so every new backup is placed on a separate tape and since your full backup fits in a single tape I suggest you leave this checkbox unchecked.

If the tape is full, does it just rewrite from the start?

Correct, if the tape is full and it's retention period is over tape job marks it as free and starts writing from the beginning.

[burak]

Hello,

It depends whether 'append' option is enabled or not. By default it is disabled so every new backup is placed on a separate tape and since your full backup fits in a single tape I suggest you leave this checkbox unchecked.

Correct, if the tape is full and it's retention period is over tape job marks it as free and starts writing from the beginning.

Hm, I don't understand this well enough. If I disabled the "append" option, the tape job will mark my full backup where the retention period is over as free and will overwrite it even without using the free space left, is this correct?

Do I have any disadvantages if I enabled "append" option? How is the scenario if append option is enabled? Is it like this:

1. Append option is enabled
2. Tape job wants to write the full backup for February on a new tape, tape is empty, it writes it at the beginning.
3. One year later, tape job writes the February monthly backup again. Tape is not empty, but append option is enabled. It writes the new full backup to the end of the last full backup?
4. Years later, tape is almost full and a new full backup won't fit. Retention for monthly backups is always 12 months, isn't it? So, each full backup on the tape is already marked as "free". It just start over and overwrite the first full backup on tape and possibly several backups because the new full backup is bigger than before.
5. One year later, tape contains |Begin|Previous year backup|2-year previous backup|...|End|. Where does the tape job starts to write?

If the tape contains empty space, why shouldn't it be used? If my backup size is small in compare to the tape size, appending monthly backups would be saved for 8 years.

[lyapkost]

If 'append' is disabled, tape is written only once with full backup and data stays on it until the retention period is over and tape job rewrites the tape.

If 'append' is enabled, your job writes February's backup from the beginnig, in a month your job writes March's backup to the same tape, to the end of February's backup etc until the tape is completely full (about 2500 Gb/300 Gb = 8 monthly backups on one tape in your case). The disadvantage is that if one wants to append monthly backups to the same tape he needs either to leave the tape in the library which may not pass security policy or to return it every month from remote storage, that is why this option is disabled by default.

Speaking about retention, it relates to the tape itself and is being updated every time the tape is written. For example, if your monthly media set has retention 12 months and 'append' is enabled, retention will be reset to 12 months every month when writing monthly backup. That's because tape cannot be erased partially so all the backups are there until retention of the newest one is over.

[burak]

Speaking about retention, it relates to the tape itself and is being updated every time the tape is written. For example, if your monthly media set has retention 12 months and 'append' is enabled, retention will be reset to 12 months every month when writing monthly backup. That's because tape cannot be erased partially so all the backups are there until retention of the newest one is over.

But I have 12 tapes for each month. If my backup takes 300GB space, I could save 8 full backups on one tape. This means, I could store February backups for 8 years on it. The problem is, after those 8 years, tape is full and have to be cleared completely to save the next backup on it. So, I could go 8 years back to each month and restore files from there but after those 8 years I lose every backup.

Is there any advantage of 'appending' backups in GFS backup scenario? Because in my case above, it wouldn't make any sense besides you buy new tapes after those 8 years.

[lyapkost]

Unfortunatelly, it's impossible to set up tape job to write only February's backups to your 'February' tape. The job 'knows nothing' about your pool of 12 tapes for every month. It can either take a new tape every time it writes ('append' disabled) or append all backups from the media set to a single tape: |Begin|February's backup|March's backup|...|September's or October's backup (partially)|End| ('append' enabled).