Backup life settings best experience

Hi,

 

I'm struggling with the backup life settings for my data sets on disk.

I can't get this to work properly; either jobs are purged from disk which shouldn't have or jobs are still on disk while they should be purged.

For example I often see that there are daily jobs on disk but without the preceding full job. So these daily jobs are useless to keep on disk.

So I'm wondering what other users have for backup life settings; based on full backup count or based on time or both, with or without force expiry.

 

My situation:

Full backup runs on Friday night = weekly backup

Incremental backup runs on Monday-Thursday night = daily backup

Backup is disk > disk > tape, used tapes are taken home each night.

 

I can divide the data on disk in 2 groups :

Group 1 - my file server: I want to be able to restore data from disk up to 2 weeks old. This server has it's own 'advanced backup options' for daily and weekly jobs.

Group 2 - all other servers: it is enough to restore data from disk up to 1 week old. These servers all use the same 'advanced backup options'.

 

Example:

On Tuesday Jan 10th, I want to restore a file from a server in group 2 from Wednesday Jan 4th.

To use the daily jobs from Wednesday Jan 4th, I would need the preceding full backup from Friday Dec 30th as well. So I would need to have 2 full backup sets on disk to be able to restore data from 1 week old,  full backup from Friday Dec 30th +  full backup from Friday Jan 6th. Correct?

 

In your opinion, what would be the correct backup life settings to accomplish this (both for the daily and weekly jobs) ?

 

Regards,

Richard

  • Hi Richard,

    Expiration requirements are dependent on more requirements than it may seem from the options description when configuring the retention in the 'Advanced Options Set'. For example, If you set both Discard After Full Backup Count and Discard After options for a Full Backup, the backup is retired only when both conditions are met. For example, if you set the Discard After Full Backup Count option to four Full Backups and the Discard After option to 30 days, the backup is retired after four Full Backup counts and 30 days.

    If any dependent Incremental or Differential Backups exist for a backup, the backup is not retired until all dependent backups have reached their retirement date.

    NOTE: When you set time-based retirement, the time component (HH:MM) is automatically set to the job save or job submit time.

    NOTE: In time-based retirement, the time component (HH:MM) does not represent the actual retirement time. It only represents the time due for backup retirement. The actual time of retirement is determined by the interval at which Media Manager scans the Media Database to identify the backups that it needs to retire. The default interval between two scans is 60 minutes. Thus, if the retirement time is set to 10:20, the backup is actually retired at 11:00.

     

    To understand this better, please see this brief section from the NetVault Backup 11.x Administration Guide under "Backup Indexes." This section will explain what you are seeing, which is basically, that retirement is reached when the retention period is met for all backups, even in an incremental sequence. Granted, the incrementals for a retired full may be rendered useless, but they still reside on disk, but the retention for the last incremental in that set must be expired before they are purged and overwritten.

     

    To answer your last question, if you were trying to restore data from Wednesday Jan. 4th, you are correct, you would need the Full from Dec. 30th. Basically, when it comes to retiring backups vs. restorability, you always want to make sure you keep retention as long as you need to restore from. So, if you have more recent restore needs, you will need to ensure you have restorability from the some means such as you have with Disk, and then have other means, such as you have with tape, for your longer retention or archival needs. If the timeframe for restore runs out on the disk copy, you need to recall from tape.

    I don't see anything wrong with how you have it currently, and you could not necessarily change it to free up space if you needed to recover data from disk from a backup over 2 weeks old. I would say the most common retention period setting people use would still be the 30 day retention at least, unless they were slim on disk based storage or something to that effect - and we do see more people use a Time Based Retention rather than Generation Based (discard after full backup count.)

  • In reply to Andre.Cooley:

    Hi,

    That's a long answer, thanks :)

    Reason I asked this is because of a SR I have about jobs that are not being purged from disk.
    So I started playing with the backup life settings to see if that would make a difference and I was wondering what settings are sort of best-practice.

    Advice from Dell was not to discard the daily jobs at all, only the fulls. The daily jobs should be deleted automatically when the full is expired (which doesn't seem to happen).
    My full's were set to discard after 3 full backups for group 1 and 2 full backups for group 2. I now changed this to 3 or 2 weeks.


    Richard
  • In reply to r.waardijk:

    Hello Richard,

    As Andre mentioned earlier, to restore data from the Inc (daily) on the 4th you would need the Full from the 30th in which it is tied to and if it's an Inc and not a diff backup you would want the Inc backups done after the 30th up to the 4th, the 2nd and 3rd of January basically, to be available to you to restore. The full run after this set on the 6th is a new set separate from the previous full on the 30th and is not needed to read from to complete this restore from the 4th.

    Group 1 you stated was setup to expire after 3 full backups.
    Group 2 you stated was setup to expire after 2 full backups.

    So let's use the Full from the 30th of Dec. as an example. If this was run when the expiry was setup to the 3 fulls it will only expire after the 3rd full is run and completed and stored.

    Having chanced the settings from 3 to 2 weeks will only apply to the jobs run after you made this change. It will not affect the Dec 30th. full job.

    So for a full detail of how the generation retirement off the Dec 30th. the full would need to expire in the following manner.
    Full stored on Dec 30th.
    Full run the following Friday Jan 6th. (1st full count).
    Full run the following Friday Jan 13th. (2nd full count).
    Full run the following Friday Jan 20th. (3rd full count is stored and triggers expiry of the Dec 30th full).
    If you look at this further you will always need available disk space to be able to hold 4 full backups with a life set to 3 full backups. Because of this we'd more often recommend using the time based retirement with days/weeks/months, etc. count as your not obligated to have as much disk storage availability.
    The date arrives and it's expired off media.

    The new retirement rules in place will have an effect over the selected 3 full or 2 week retirement of backups. They are the following;

    Retire all backups when the last backup is retired.
    Retire all backups when the first backup is retired.
    As Andre mentioned these options are explained in our latest 11.2 Administrator guide starting on page 82.

    2 weeks expiration;
    Placing the current full for 2 weeks backup life and it's associated incs with no life upon them, using the retirement rule of, "Retire all backups when the first backup is retired", should clear each of your fulls along with associated incs out every 14 days.

    I hope this helps!

    Regards,

    Scott K.
  • In reply to Scott.Kline:

    Hi,

    @Scott: I think you've been working on my SR as well concerning jobs not being purged from disk.
    This topic is a bit related to that issue.


    You mention the new retirement rules....well, I still have to find them.
    The only retirement timing options I can find are here:
    Change Settings > Server Settings > Media Manager > Retirement Timing Control

    But there is nothing like "Retire all backups when the last backup is retired" or "Retire all backups when the first backup is retired".

    The only available options are:
    - Force Always
    - Keep Duplicates with Dependents
    - Allow Duplicates with Dependents to Retire


    Current backup life settings:
    Daily (incrementals) : never discard
    Weekly (full) : discard after 2 weeks, force expiry

    Current retirement timing settings:
    Allow Duplicates with Dependents to Retire


    But with these settings my daily jobs are not purged from disk. So for every server there are still daily jobs on disk without the previous full backup using unnecessary disk space.
    The only way to get rid of them is to delete the savesets manually.
  • In reply to r.waardijk:

    Hello Richard,

    I understand the confusion.

    The options "Retire all backups when the last backup is retired" or "Retire all backups when the first backup is retired" in 11.x are located within the Advanced Options section of the backup.

    It's described under the Force check area.
    Checking Force = retire full and all Incr when the full expires, (as traditionally seen in our 9.2 and below versions). Or <Retire all backups when the first backup is retired>.

    Un-check Force and full will stay around even after its set expiry, thus waiting for Incs to expire. <Retire all backups when the last backup is retired>.

    Go to > Manage Sets>Set Type> Advanced Backup Options>Select any Backup option recently created on 11.x>Choose "Edit">Choose Backup Life section,
    This is on a per job basis.

    You mentioned this earlier within the "Change Settings > Server Settings > Media Manager > Retirement Timing Control", and there is a pull down menu to choose "Force Always ". Which would be the option for Global settings, (every job), using the "Retire all backups when the first backup is retired".

    Thanks,

    Scott K.