Backup Job has Too Many Restore Points - Considerations and Causes

Actual Too many restore points

When using Forward Incremental Retention, creating a Full restore point is what starts a new backup chain. Retention is enforced when the entirety of the oldest backup chain is outside retention. Therefore, if a new Full fails to be created, the backup chain will become longer as more increments are created, and retention cannot be enforced as expected. For this reason, the most common situation where there are legitimately too many restore points occurs when the scheduled Synthetic or Active full fails to be created.

Remember, until all restore points of the oldest chain have fallen outside of retention, no portion of that chain can be deleted. Furthermore, if there is only one backup chain on disk because a second full could not be created, retention will never be enforced.

If, due to repository space restrictions, it will never be possible to create enough full restore points on disk to allow retention to function correctly, either (a) reduce retention or (b) switch to Forever Forward Incremental.

Scenarios that lead to Full restore point creation failure:

• Scenario 1: Ignored failure
  
  This often occurs when a backup job is configured to create a Synthetic Full on Saturdays, and each Saturday, the Full fails to be created. The job reports a failure for the Saturday run, but on the next day, the job completes successfully. The Saturday's job failure may be overlooked because it is followed by a successful run, creating the impression that whatever went wrong fixed itself the next day.
  
  Recommended Action:
  Investigate why the Full failed to be created.
  
  
• Scenario 2: Disabled Job
  
  The backup job was disabled on the day that the Full restore point was scheduled.
  
  Recommended Action:
  Ensure that the job is not disabled for the next day that a Full is scheduled to be created.
  
  
• Scenario 3: Job Overrun
  
  An Incremental session ran for more than 24 hours, causing the next job run to not occur on the day a Full is scheduled to be created.
  
  For example, the job is scheduled to run every day at 10 PM. The incremental job started at 10 PM on Friday but did not complete until 2 AM on Sunday. Because the Incremental run of the job was still running at 10 PM on Saturday, the job run that should have created the Full on Saturday was not able to trigger. The next run of the job, at 10 PM on Sunday, does not create a Full because that is not the day a Full restore point is scheduled.
  
  Recommended action:
  Investigate why the incremental job run took so long. If the incremental job runs often take longer than 24 hours, consider ways to improve job performance to ensure that the job completes within 24 hours or consider reducing the number of objects in the job.
  
  
• Scenario 4: Job Chaining Issue
  
  There are multiple backup jobs, and they are scheduled to run sequentially using the After this job scheduling option. The first few jobs in the sequence run and create their Full restore point, but in doing so, they take so long that they push the later jobs in the series to start on the next day of the week. Since those later jobs began on a different day of the week than was scheduled to create a new Full restore point, they do not create a Full restore point.
  
  For example, ten jobs are configured to run in sequence, the first job starts the sequence by running daily at 9 PM, and all the jobs are set to create a Full restore point on Saturday. On Saturday, as each job in the sequence starts and creates its respective full, the later jobs are pushed to start running past midnight, meaning they technically start on Sunday and therefore did not create a new Full.
  
  Recommended action:
  Generally speaking, it is not advisable to use Job Chaining when jobs are set to create Fulls on specific days of the week. Instead, the better option is to fine-tune a combination of task limitations, job priorities, and staggering job start times. In most environments, multiple jobs may be set to start simultaneously, and the task limitations on proxy and repository will control how many tasks occur simultaneously to ensure that the environment is not overwhelmed.
  
  If, however, there remains a desire to use Job Chaining, set the first job that starts the chain of jobs to start at Midnight. Doing so will ensure that the first and all jobs after it run on the same day, assuming the last job in the chain begins before the next day.