Roll up job fail with error: "Cannot release a buffer that has already been released" (4042284)

Roll up job failed with error message: "Cannot release a buffer that has already been released".

Logs might include a message indicating an invalid record found on dvm increment refs of dfs_record entry 0x#######

Option 1 - If there is an archive or replication of the affected system

1) Delete the backup chain with the corrupted reference

2) Allow deferred delete jobs to complete

3) Replicate or import the recovery points back into the system

Option 2 - Create New Agent ID

If the customer still have plenty of space in their repository the recovery points of the agent with the issues can be deleted and sent to recovery points only, create a new agent ID on the agent registry and protect the agent again to perform a new base image. And leave the recovery points only with the issue for a time and then delete them to recover that space.

1) Remove the affected agent from the Core console without removing the recovery points as they will remain as recovery points only

2) Navigate to the affected agent server and stop the Rapid Recovery agent service

3) Open the Regedit and locate the key HKEY_LOCAL_MACHINE\SOFTWARE\AppRecovery

4) Rename the AppRecovery to AppRecoveryOLD

5) Start the services of Rapid Recovery agent service, A new agent ID will be created

6) Protect the agent again in the Core (Protect Machine)

7) Perform a new base image

Option 3

Run RRT to remove corrupt record
 

About RRT

The RRT tool is a corruption removal tool, not a corruption repair tool. It scans and deletes corrupt volume images. Volume images are the individual volumes (i.e. drive letters) being backed up per agent that make up each recovery point. For each corrupt volume image that is removed all subsequent images (i.e. incremental snapshots) for the same volume in the recovery chain will be removed as well as each volume image relies on the one before it. Thus if all volume images on one recovery point are corrupt/deleted then the remaining recovery points will be removed up until the next base image.  
 
When the RRT tool is running, all activities on the core must be paused.  The following jobs will not be available during this time: backups, replication, exports, rollups, and restores.
 
The repository recovery tool requires free space on a local volume to the Core Server equal to 2% of the size of the repository.  If this amount of free space is not available, we cannot guarantee the tool will complete successfully.  Please note there is no guarantee that the repository will be recoverable.  We will need to monitor the repository for a week’s time after RRT is run to make sure all the corruption was detected and removed.

Please note, this tool can only be run in the presence of Support Engineer.